Industrial Gases

Industrial Gases
!"#$ &'() #*+)*+#,*'--. -)/+ 0-'*1

Industrial Gases
N.A. Downie
Partner
Maran & Co.
Guildford, UK
KLUWER ACADEMIC PUBLISHERS
NEW YORK, BOSTON, DORDRECHT, LONDON, MOSCOW
eBook ISBN:
Print ISBN: 0-751-40352-0
2002 Kluwer Academic Publishers
New York, Boston, Dordrecht, London, Moscow
Print 1997 Chapman & Hall
All rights reserved
No part of this eBook may be reproduced or transmitted in any form or by any means, electronic,
mechanical, recording, or otherwise, without written consent from the Publisher
Created in the United States of America
Visit Kluwer Online at: http://kluweronline.com
and Kluwer's eBookstore at: http://ebooks.kluweronline.com
Preface
The natural world is a wonderful place but much of the natural world would
be inaccessible or invisible to us without the products of industry; for ex-
ample air and road travel, underwater breathing apparatus and television. It
is not simply the products of industry, however, that can enrich our lives; the
activities and techniques of industry can, too.
The story of gases in industry is just one of a hundred interlinked subjects
needed to describe the totality of modern industry. The subjects of this
book, nitrogen, oxygen, carbon dioxide, argon and others, are inextricably
woven into the fabric of modern manufacturing. Almost no branch of indus-
trial processing is untouched by them. Gases begin their role close to pri-
mary extractive industries, for example in the processing of ore. The
manufacture of materials from those as common as steel to such exotics as
semiconductor-grade indium phosphide is now dependent on gases. The
conversion of those materials into useful products and services involves a
wide range of applications of gases. Gases now affect almost all fields of
human activity: what we eat, what we drink, how we travel, what we read,
our health and environment.
This book is about the industrial gases that have a key role in general in-
dustry. Methane or propane gas used for heating purposes could be easily
replaced by other heat sources. Other gases, such as ethylene, are inter-
mediates in chemical manufacture and exist only fleetingly in pipes deep
within specialised chemical processing complexes. Although these are men-
tioned and described, the main thrust of enquiry will be directed towards
those gases which are not simply burnt but which are visibly produced and
used in industry at large.
Nitrogen, the biggest industrial gas worldwide, is today still mainly made
by liquefying and then distilling air. This process is now very much more ef-
ficient than in former times. There remain, however, the intrinsic inefficien-
cies attached to cooling air to 195C to liquefy it prior to separation.
Despite many improvements in this process, and the considerable advan-
tages of scaling, it continues to require many times the thermodynamic
separation energy. Hence there are now important challenges being made to
cryogenic separation. Pressure swing absorption units have been practical
for many years and are now edging into the marketplace, although their
vi PREFACE
thermodynamic efficiency is still not high. An even bigger challenger may be
waiting in the wings: membrane separation. Membranes are probably the
simplest possible way of separating gases: a filter passes the molecules
required and sieves out the ones not required. Although a process as simple
as this is possible, it is not yet truly practical and economic. These and other
promising new processes for making gases will steadily grow more important
as new applications for gases put new demands on production techniques.
New applications are constantly being found for gases and the whole
balance of gas production has been shifted in recent years. Chapter 2
examines how nitrogen has now outstripped oxygen in volume terms in
many countries. Although the steel industry and chemicals still take vast
volumes, newer areas of application have opened up. The rapid freezing of
foodstuffs with liquid nitrogen is a large and fast-growing sector, and the
semiconductor industry has seen a meteoric rise in demand, becoming the
most important consumer of gases in some countries. The future will also
hold big changes, and Chapter 4 looks at what new applications may become
important and how these will shape the industry and the technology used. A
glossary of words that are used in a particular sense in the gas industry, or are
unique to it, is included at the end of the book.
There is a lot of sophistication in the gases industry. Despite this, its basis
is still firmly grounded in the kind of physics that students learn early in
university or, in the UK, in the final years of a good school. Such students
have, I believe, all the necessary scientific tools to take on a whole field of
industrial science. They will find from this book that the part of industrial
science that relates to industrial gases is easily comprehensible. A firm basic
grounding in science is, however, not quite sufficient. There are many
simple facts that are very useful to know for practical purposes but are
difficult to fit into a systematic, scientific frame. It is impossible to give a
comprehensive encyclopaedia of these in a short and wide-ranging book like
this. Nevertheless, I have included a fair number of hints, tips and anecdotes
that fall into this category, most from my own personal experience, which I
hope will be useful.
The industrial gases industry is now one of the largest in the world with
sales of $20bn in 1992. More than a million tonnes of industrial gases are
used every day. Gases have put men on the moon and helped to uncover the
secrets of the atomic nucleus. They are used in almost every conceivable
sector of industry in every corner of the world. Despite this, gases are largely
a hidden industry. Only rarely are gases directly used. As a result, relatively
few people, even within the quarter of a million who work in the industry,
know much about them. The secrets of how oxygen is made by distilling
liquid air, for example, or how to make diamonds from a gas mixture
deserve a wider audience. I know that I have enjoyed uncovering these
secrets over the years and I hope that this book will enable the reader to
appreciate them, too.
Acknowledgements
The author is indebted to a large number of people for their help in the
preparation of this wide-ranging book starting with my forbearing family
and my partners in Maran & Co. Space does not permit me to list everyone,
but I should at least list the organisations from which they came. (Please
note, however, that responsibility for the opinions and facts given are
entirely mine.) Space also forbade the use of all the illustrations and other
information that people kindly provided; please accept my apologies if your
contribution was not used.
Accles & Pollock
AFROXCompany
AGA Company
Air Liquide
Air Products
AIRCO Company
ARC Machines
Atlas Copco
Balzers-Pfeiffer
Bergbau Forschung
Birwelco
BOC Group
British Meteorological Office
British Steel Corporation
British Technology Group
Broeder Mueller Company
Cambridge Fluid Systems
Cameron Balloons
CEA-LETI R&D institute,
Grenoble, France
Ceodeux Company
CERN R&D institute, Geneva
CHAM
Chesterfield
Courtaulds
Cryoplant
DESY R&D Institute
DOW Chemical
DRAX Torches
DuPont
Edwards High Vacuum
Epichem
Eurometers
FLUOR Corp.
Gardner Cryogenics
GEC Semiconductors
Guildford Museum
Hale Hamilton Valves
Hawker Siddeley Switchgear
High Energy Physics Group,
Imperial College, London
Hoechst
HYT Corp
IBM
ICI
IMEC R&D Institute, Leuven,
Belgium
IMI Marston
Institute for Aviation Medicine,
Farnborough
viii ACKNOWLEDGEMENTS
JEMI UK
JESSI Semiconductor R&D
Programme, Munich
Linde
Liquid Air
Luxfer
Madder Scientific/Systech
Malaysia Oxygen Company
Matheson
Messer Griesheim
Millipore
Monsanto
Nippon Sanso
Ohmeda
Oxford Plasma Technology
PALL Filter
Pilkington
PRAXAIR
Puritan-Bennett
SAES
Siemens
Sulzer Brothers
Swagelok
Tetronics
Texas Instruments
TI Chesterfield
Tohoku University, Japan
Union Carbide Corp.
University of Edinburgh
University of St Andrews
University of Surrey
US Steel Corp.
VG Instruments
Wacker
Wormald International
Contents
1 An introduction to gases and the gases industry
1.1
1.2
1.3
The discovery of gases
Early use of gases in industry
1.2.1
1.2.2
1.2.3
Coal gas
Oxygen and acetylene
Oxygen for steel
Properties of gases
1.3.1
1.3.2
1.3.3
1.3.4
1.3.5
1.3.6
1.3.7
1.3.8
1.3.9
1.3.10
1.3.11
1.3.12
1.3.13
1.3.14
1.3.15
1.3.16
Ideal gases and the energy stored in a compressed
gas
Real gases: deviants from the ideal
The finite size of gas molecules
Heat capacities
Bernoulli and viscous gas flow
Poiseuilles equation
Turbulent and laminar flow
Compressible flow
Computational fluid dynamics
Diffusion
Molecular flow: small orifices and vacuum systems
Heat transfer in gases
Thermal conductivity and viscoscity of gases
Vapour pressure of liquids
Sublimation and supercritical fluids
Dissolved gases in liquids
1.4
1.5
1.6
1.7
Some specific gases
1.4.1 Special gases
The gases industry
1.5.1
Which gases matter?
Corporate structure of the industrial gases industry
1.6.1
1.6.2
1.6.3
1.6.4
A typical industrial gas company
Monopoly in the gas industry
Gases industry supply chain
Geography of gas supply
Financial features of the gases industry
1
3
3
4
5
5
7
9
11
13
14
15
15
19
22
23
26
26
28
32
34
37
41
42
42
44
45
47
47
48
51
1
11
x CONTENTS
1.7.1
1.7.2
1.7.3
Product exchange in special gases business
Take-or-pay and argon credits
Scaling-up gas plants
1.8 The key role of gases in industry
1.8.1 Smokestack and leapfrog economies
2 Gas technology
2.1 Separating air gases
2.1.1
2.1.2
2.1.3
2.1.4
2.1.5
2.1.6
2.1.7
Introduction
Gas plant building blocks
Compressors
Thermal transfer in gases: heat exchangers
Distillation of air
Pressure swing absorption
Membrane separation of gases
2.2 Making other gases
2.2.1
2.2.2
2.2.3
2.2.4
2.2.5
2.2.6
2.2.7
2.2.8
2.2.9
2.2.10
2.2.11
2.2.12
Gases from chemicals
Carbon dioxide
Acetylene, ethylene and unsaturated hydro-
carbons
Fuel gases: methane, propane and butane
Hydrogen
Chlorine and fluorine
Ozone generators
Helium
Nitrous oxide
Production of special gases
Purifiers
Preparation of gas mixtures
2.3 Gas measurement and analysis
2.3.1
2.3.2
2.3.3
2.3.4
2.3.5
2.3.6
2.3.7
2.3.8
Gas flow measurement
Pressure measurements
Analysing gases
Gas chromatographs
Infrared and related analysers
Mass spectrometry
Other analytical principles
Monitoring air for gases
2.4 Gas handling and storage
2.4.1
2.4.2
2.4.3
2.4.4
2.4.5
2.4.6
Pipelines and pipeline components for gases
Valves and other pipeline components for gases
Removing particulates: filters and cyclones
Gas flow control systems
Liquid gas storage and handling
Storing gas in gasometers
52
52
53
57
59
61
61
61
63
64
78
83
98
111
117
118
119
120
124
125
129
130
132
133
134
137
144
147
147
159
162
164
172
176
181
188
191
192
203
213
218
225
235
CONTENTS xi
2.4.7
2.4.8
2.4.9
2.4.10
2.4.11
Making and using compressed gas cylinders
Transport of gases
On-site generation of gas
Gas losses
Vacuum pump technology
2.5 Safety and environmental issues
2.5.1
2.5.2
2.5.3
2.5.4
2.5.5
2.5.6
2.5.7
2.5.8
2.5.9
2.5.10
2.5.11
2.5.12
2.5.13
2.5.14
Where does gas go in an accident?
Toxic and asphyxiant gases
Cryogenics
Flammability
Flammability limits
Ignition energy
Oxygen index
Safe working in flammable atmospheres
Gas release accidents
Safe equipment design
Safety audit and HAZOP/HAZAN analysis
Environmental issues
Industrial gases and the environment
Scrubbing
3 Applications of gases in industry
3.1 Gases in metal smelting and extraction
3.1.1
3.1.2
3.1.3
3.1.4
3.1.5
3.1.6
3.1.7
3.1.8
3.1.9
Oxygen in ferrous metals
Blast furnaces and the smelting of iron
Increasing prospects for oxygen in blast furnaces
Conversion of iron to steel
Basic oxygen and LindDonawitz oxygen-lancing
steelmaking processes
Mini-steelworks and arc furnace steelmaking
Argon for stirring
Oxygen non-ferrous metal extraction
Oxygen flame enhancement
3.2 Gases in metal fabrication
3.2.1
3.2.2
3.2.3
3.2.4
3.2.5
3.2.6
3.2.7
3.2.8
3.2.9
Gases in metal casting
Furnace atmospheres for metal processing and
heat treatment
Flame welding and cutting
Flame spraying/coating
Arc welding and cutting
Laser welding and cutting
Ceramic and metal powder forming technology
Production of ferrites and ceramic high critical
temperature superconductors
Hot isostatic pressing
236
252
254
255
256
262
262
263
265
266
269
271
271
272
273
279
285
288
293
294
298
298
298
298
301
303
304
305
306
307
308
310
310
311
316
322
325
332
333
334
336
xii CONTENTS
3.2.10 Metal powders by spraying
3.3
3.4
3.5
3.6
3.7
Industrial gases and chemical manufacture
3.3.1
3.3.2
3.3.3
3.3.4
3.3.5
3.3.6
3.3.7
3.3.8
Oxygen for catalytic converters
Gasification of coal
Nitrogen and injection in oil wells
Oxygen for oxidation of chemical intermediates
Oxygen-enhanced sulphur recovery
Ammonia and fertilisers
Inerting chemicals
Miscellaneous chemical applications
Gases in the food industry
3.4.1
3.4.2
3.4.3
3.4.4
3.4.5
3.4.6
3.4.7
3.4.8
Freezing and chilling
Dry-ice freezing
Gases as preservative chemicals
Carbonated drinks
Carbon dioxide in intensive agriculture
Gases in pest control
Margarine and edible oils
Decaffeinated coffee and detarred tobacco
Gases in the semiconductor industry
3.5.1
3.5.2
3.5.3
3.5.4
3.5.5
3.5.6
3.5.7
3.5.8
3.5.9
3.5.10
3.5.11
3.5.12
3.5.13
3.5.14
3.5.15
The need for high purity in semiconductors
Making ultraclean gases
Keeping ultraclean gases ultraclean
Ultrapure gas pipework simulation
Further developments in ultraclean gas technology
Semiconductor deposition processes
Plasma etching: a key technology
Other plasma processes
Semiconductor effluent gases
Other processes in the wafer fab
Gases in IC assembly and hybrid circuits
Gases in compound semiconductor manufacture
Gases for optical fibres
Gases in general electronics
Thermionic valves
Gases in water treatment
3.6.1
3.6.2
3.6.3
3.6.4
3.6.5
3.6.6
Oxygen for controlling aerobic sewage treatment
Drinking water: treatment
Industrial wastewater treatment
Control of pH with
Gas stripping
Oxidation treatment
Industrial cryogenics
3.7.1 Properties of liquefied gases
338
339
339
340
341
342
343
344
345
348
351
351
359
360
362
365
366
368
370
371
373
377
379
386
387
392
401
408
411
413
416
420
422
425
427
429
431
436
438
438
439
440
442
442
CONTENTS xiii
3.7.2
3.7.3
3.7.4
3.7.5
3.7.6
3.7.7
Cryogenic phenomena
Cryogenic practice
Dry ice bead-blasting
Liquid helium and NMR/MRI
Gaps in the cryogenic spectrum
Cryogenic pipelines
3.8
3.9
Gases across the breadth of industry
3.8.1
3.8.2
3.8.3
3.8.4
3.8.5
3.8.6
3.8.7
3.8.8
3.8.9
3.8.10
3.8.11
3.8.12
3.8.13
Gases in space
Cleaning: thermal blasting of stone
Firefighting
Gases in tyres and inflatables
Gases in glass manufacture
Coated glass
Gases in double glazing
Gases in ceramic firing
Electrical power generation
Gases in lighting
Balloons and airships
Aerosol spray cans
UV coatings
Medicine and science
3.9.1
3.9.2
3.9.3
3.9.4
3.9.5
Gases for medicine
Measurement of lung function with gas mixtures
Other medical applications of gases
Breathing mixtures: diving and aviation
Gases: vital tools of research
4 Future of industrial gases
4.1
4.2
4.3
Market pull
Prospects for new production processes
4.2.1
4.2.2
4.2.3
4.2.4
4.2.5
4.2.6
4.2.7
Bulk nitrogen and oxygen
Pressure swing absorption
Membrane separation
Liquid hydrogen
More efficient gas cylinders
Carbon dioxide
Acetylene
Developing new applications
4.3.1
4.3.2
4.3.3
4.3.4
4.3.5
4.3.6
Oilfield inspection
Transonic wind tunnels
New application for liquid nitrogen freezing
Supercritical carbon dioxide
New developments in steel production
Changes in the liquid fuels market
442
450
467
468
470
471
472
472
474
474
475
476
478
482
483
485
487
492
496
500
502
502
509
510
512
514
518
518
519
520
521
521
523
524
525
525
525
526
526
527
527
528
529
xiv CONTENTS
4.3.7
4.3.8
4.3.9
4.3.10
4.3.11
Nitrogen tunnels for semiconductors
HT
c
superconductor cooling
Diamond production
Production of superstrong whisker materials
Greener processes
4.4 Epilogue
Appendix A: Glossary
Gas industry terms
General terminology
Tradenames
Appendix B: The human nose as a gas detector
Appendix C: Use of industrial gases in schools
Appendix D: Gas cylinder colour marking
References
Further reading
Index
529
531
532
533
534
535
537
537
540
542
543
547
553
555
557
559
1 An introduction to gases and the gases industry
1.1 The discovery of gases
A gas is the simplest form of matter. In the kinetic theory of matter, a gas is
simply a group of freely moving small spheres whose only interaction with
each other and other matter is via elastic collisions. It is in effect matter
having neither constant shape nor volume, and substantially invisible and
intangible.
An invisible and intangible form of matter seems at first to be an unlikely
concept. However, it arose surprisingly early in history. The concept was
already well established in classical Greek thought. That the Greeks
recognised air as an actual substance distinct from vacuum is clear since they
talk about objects such as clouds, water or the earth being formed by the
condensation of the air (Wightman, 1950). Perhaps analogy between the
behaviour of gases and the behaviour of (readily observable) liquids helped
understanding, along with the observation of the visible behaviour of
smoke- or mist-laden air.
The use of air-handling machinery, for example bellows, was also known
to the Greeks. Perhaps as far back as the 2nd century BC the Greeks had
developed musical organs driven by bellows. Being experts with mechan-
isms using water developed for their clepsydra or water clocks, it is thought
that the Greeks also devised pipe organs with ingenious hydraulus
mechanisms employing water in a cistern to stabilise the air pressure
delivered (Elvin, 1971).
The need for intense heat in the early metals, ceramic and glassware
industries led to knowledge of the use of low-pressure air in furnaces, craft
knowledge that was passed on down the centuries. Alchemists steadily
advanced the state of practical gases technology and by medieval times
several different gases were recognised in addition to air (although at this
stage only air had any industrial importance). Illustrations in some of the
very first printed books, showing bellows, valves and pipework, give an idea
of the state of the technology after the Middle Ages (Agricola, 1556).
Lacking a scientific approach, however, and without suitable instruments,
gas-handling equipment or basic scientific knowledge, progress was slow.
Major advances were at last made in studying the properties of air in the
1600s. Robert Boyle, in 1662, published experiments with a mercury
barometer which quantified the change of pressure of air with volume.
2 INDUSTRIAL GASES
However, accurate experiments on the change of pressure of air with
temperature had to wait for Gay-Lussac in 1802.
More rapid progress in understanding gases came with the rise of chemistry
as a science. Joseph Black might be given credit for making advances in this
direction in 1755. He published experiments which recognised at least two
gases as distinct, ponderable substances, rather than modified air: the
oxygen in air and the most frequent combustion product, However,
more profound discoveries were not possible until improvements were made
in gas handling by Priestley and others from around 1770 onwards.
All the early chemists made important discoveries in gases around this time
(Ackroyd, 1935; Lavoisier, 1952). Priestley alone, for example, could claim
the discovery, or at least the first correct identification, of the oxides of
nitrogen, HCl, ammonia and CO. In 1789 Lavoisier wrote his
systematic Elementary Treatise on Chemistry which is recognisably a modern
chemistry book and dedicates a large proportion of its text and illustrations to
gases and gas handling, which was by then already well-developed. By the
turn of the century oxygen, nitrogen, HCl, HBr, and many other gases
were all fully described and routinely made in chemical laboratories.
Three of the best-known chemical pioneers Priestley, Scheele and
Lavoisier could each claim to have been the discoverer of oxygen, around
1775. The use of the reversible chemical reaction of mercury with oxygen
(Gibbs, 1967; Anon, 1980) pointed the way to the first commercial oxygen
separation using barium. To Lavoisier must go the credit for providing a basic
list of irreducible substances (chemical elements) and the naming of oxygen.
Oxygen means acid producer in Greek. Oxygen was the first gas, other than
air, to be used on a large scale in industry, although inflammable air
(hydrogen) had been discovered in 1767 by Cavendish.
The advance of the atomic theory of chemistry, led by John Dalton,
enabled the chemical formulae of many gases to be identified correctly by the
1800s. This in turn fed back to chemistry because Avogadro had predicted
that the density of any gas would be precisely proportional to its molecular
weight. The use of Avogadros concept thus enabled the molecular formulae
of gases and, eventually, the atomic weights of all elements to be precisely
determined. However, although Avogadro first described his molecular
theory of gases in 1811, it was half a century or so later before the theory was
accepted.
Eventually a virtuous circle became established in which one scientific
discovery led to another or to another technical development which in turn led
to another scientific discovery. By 1860 the physics and chemistry of the
majority of industrial gases had been worked out. For example, the concept
of a perfect or ideal gas one in which the molecules are spherical and
undergo only simple elastic collisions according to kinetic theory was
introduced in this era of history and the ideal gas equation formulated. This
neatly encapsulated the formulae of Boyle, Gay-Lussac and Avogadro, and
simplified the kind of calculations needed in industry.
AN INTRODUCTION TO GASES AND THE GASES INDUSTRY 3
In parallel with this scientific development were technical developments
due largely to the requirements of the rising steampower industry: by the
mid-nineteenth century gases could be handled not just at near-atmospheric
pressure but at low pressure or high pressure up to 10 bar (e.g. Anon, 1989).
Helium and argon were discovered late in the nineteenth century.
Terrestrial helium was discovered by Hildebrand in 1891 (Anon, 1995)
through its optical emission spectrum, the helium being formed extremely
slowly by alpha particle emission in minerals containing radioactive
elements such as uranium. Helium had ingeniously been predicted more
than 20 years earlier by detecting its emission spectrum in the sun. Lord
Rayleigh observed as early as 1882 the slight differences in the density of
gases prepared by different routes. Work by Sir William Ramsay established
argon in 1894, preparation being by reacting combusted air (i.e. nitrogen +
inerts) with magnesium. Ramsay went on to discover other inert gases.
Air distillation and the technology of cryogenics, which were being
developed at the end of the nineteenth century, allowed the discovery of
neon and the heavier rare gases, krypton and xenon: an example of a
virtuous circle operating. One science (low temperatures) led to a tech-
nology (air distillation) which led to another science (rare gases). Neon was
found in the vapour above an early sample of liquid argon, whilst krypton
was found by its optical emission spectrum. Xenon was found soon
afterwards by distilling the crude krypton. By 1898 all the stable rare gases
had been discovered. The considerable scientific feat that this represents
should perhaps be emphasised; the discovery of part per million components
in the air, when those components are undetectable to human senses and
completely inert chemically, must rank as a major achievement.
By this point in history many other gases had not been discovered.
However, the remaining undiscovered gases, with one or two important
exceptions,
1
are not important in industry today. The stage was now set for
industrial gases to blossom into a major industry on a firm scientific and
technical footing.
1.2 Early use of gases in industry
1.2.1 Coal gas
Coal gas constituted the first major industrial use of gas: as early as 1600 coal
was being heated in retorts and the gas given off burnt. In fact, the gas was a
1
One class of gases that the Victorian chemists totally failed to discover was the chlorofluor-
carbons, which had to wait until Midgelys work in the 1930s. Midgely, whilst working for
General Electric, discovered that, despite being made from some of the most reactive elements
known, the CFCs had an extraordinary level of inertness.
4 INDUSTRIAL GASES
by-product of the requirement for coke for iron smelting; UK charcoal
supplies being exhausted (Williams, 1981; Buchanan, 1992).
At first the gas was very impure and used only on a small scale. The basic
components, hydrogen and carbon monoxide, were adulterated with
ammonia, sulphur compounds and hydrocarbons even as heavy as benzene.
However, great strides were swiftly taken in the technology and then in the
distribution and use of coal or town gas. By 1800 a major industry had arisen:
heating coal to make coal gas which was then used for lighting. The later
invention of the gas mantle improved the efficiency of gas lighting and
extended the technical lifetime of the technique well into the twentieth
century. The mantle was actually a thorium oxide-coated grid which when
heated emitted a slightly greenish white light.
Many other early industrial processes also evolved gases as by-products
but most of these were unpleasant or even dangerous rather than immedi-
ately useful. For example, the biggest Victorian era chemical industry, the
alkali/soap industry, relied at first on the infamous LeBlanc process. This
involved adding sulphuric acid to salt, leading to the emission of copious
amounts of hydrogen chloride (heating with limestone and charcoal
completed the conversion into crude sodium carbonate and produced
another pollutant, a calcium sulphide sludge). After serious pollution
incidents, the worlds first anti-pollution body, the UK Alkali Inspectorate,
was set up in 1863 (e.g. Anon, 1993).
1.2.2 Oxygen and acetylene
Limelight was one of the first applications for an industrial gas, in this case
oxygen. An oxygen-intensified gas flame could be used to heat rods of
calcium oxide (lime) to very high temperatures. At these temperatures, in
excess of those achievable with air alone, the lime gave off a powerful white
light far superior to that obtained using coal gas. Limelight, as the system
was known, was much more satisfactory for illuminating large halls, such as
theatres, than oil or gas lamps.
Limelight used oxygen derived from sources such as the barium oxide
Brin process. This process was based on the production of barium peroxide
by roasting barium oxide in air at 590C, then raising the temperature to
870C. At 870C the peroxide formed decomposes back into oxide, releasing
more or less pure oxygen which can then be cooled and compressed into
steel gas cylinders. Although crude, the process was ingenious in that it
required no continuous input of raw materials other than air and energy. Its
active process was one of reversible absorption, making it superior to an
irreversible process such as heating of nitrates.
At the turn of the century in Germany, Carl von Linde developed the
liquid air distillation process for oxygen. This so revolutionised the
production of oxygen that it was quickly adopted elsewhere, for example in
the UK and the USA. His work was paralleled by that of Georges Claude in
France. Both men founded companies which are today large and successful
industrial gases firms: the Linde company of Germany and the USA, and the
LAir Liquide company of France. Now, at last, oxygen was available in
large quantities at low prices.
Meanwhile another gas business had been growing up around acetylene.
Gas lamps, particularly those with an incandescent mantle, are an effective
source of light. However, the town gas of the day could only be provided via
a pipe from the gasworks, and thus portable gas lighting and lighting outside
of towns was impossible. The development of the calcium carbide/acetylene
lamp was therefore an important innovation (Figure 1.1). Acetylene burns
with a flame which is intrinsically incandescent, much more so than any
other hydrocarbon, and so is eminently suitable for lighting. Portability was
assured because acetylene can be derived simply by dripping water onto
calcium carbide chips.
Oxy/acetylene arose out of the combination of these two Victorian gas
industries, aided by the discovery that acetylene could be safely stored in
large quantities in cylinders once it had been dissolved into an acetone
absorbed on a porous inert packing. It caught on rapidly after the First
World War and delivery networks of oxygen and acetylene cylinders were
built up all over the world. On the back of this business most of todays major
gas companies grew to be large and multinational corporations: companies
such as BOC, LAir Liquide, Union Carbide and Linde (then in both the US
and Germany).
1.2.3 Oxygen for steel
In the intense rearmament era after the First World War, demand for steel
was intense. Linde in Germany realised that oxygen could enhance the
productivity of steel plants and devised processes for Bessemer converters.
These processes required enormous quantities of oxygen, much larger
volumes than had been produced hitherto, and resulted in rapid improve-
ments in the efficiency and scale of liquid air plants. Plants in the range of a
few tens of tonnes per day became quite common. Driven by this immense
demand for its products, industrial gas businesses began to expand rapidly
and to resemble todays huge industry.
1.3 Properties of gases
Gases are studied by physicists interested in kinetic theory, by chemists
interested in their reactions, and by mechanical engineers interested in
engines. Standard textbooks provide useful subsets of gases science and
know-how aimed at satisfying the study needs of these groups. Many good
6 INDUSTRIAL GASES
Figure 1.1 Acetylene bicyle lamp, c. 1900 (courtesy of Matthew Alexander, Guildford
Museum). a, b: alternative burners; c: rotating body allowing instant burner change; d: union
nut attaching burner.
standard works are available (see Further Reading) and the sections in this
chapter are intended only as reminders of physical principles. The standard
works should be consulted for details. Moore (1978) on physical chemistry,
together with Flowers and Mendoza (1970) on kinetic theory, Zemanski
(1968) on thermodynamics and Douglas et al. (1984) on flow are good
introductions. A technical manual on chemical engineering, like Perry
(1984) is essential for accurate, detailed calculations, as are compilations of
data such as are included in the CRC Handbook of Chemistry and Physics,
Tables of Physical and Chemical Constants (Kaye and Laby, 1995) and
specialist gas data publications.
1.3.1 Ideal gases and the energy stored in a compressed gas
The ideal gas equation describes the relationship between the equilibrium
pressure (P) exerted by a known mass of ideal gas confined in volume V at
temperature T
where n is the number of gram-molecules of gas, R is 8.31 J (the gas
constant) and T is the absolute temperature in degrees Kelvin.
The ideal gas equation approximately describes the behaviour of many
gases at ordinary temperatures and pressures, and is still today enormously
useful in industrial calculations.
Using the ideal gas equation, the energy (E) stored in a compressed gas
can be calculated
where and are final pressure and volume, and and are initial
pressure and volume. The energy stored in compressed gases is only
occasionally apparent in everyday life. There are some gadgets that literally
replace springs with gas. Examples of this are gas struts, actually gas
confined within a piston and cylinder, which use pressures of 50 bar or so to
support swing doors, e.g. the tailgate of a car.
The energy that can be stored in compressed gas can be released very
suddenly with immense, potentially explosive, power, something which
should always be borne in mind when designing gas systems and com-
ponents. This is demonstrated routinely in the testing of gas cylinders
(Figure 1.2). Occasionally a cylinder is tested to destruction by pumping it
up to twice, or more, its rated pressure. The result is a very violent explosion
which twists the very tough cylinder into contorted shapes and may blow
pieces a considerable distance. The energy stored in a standard cylinder of
at, say, 400 bar pressure is of the order
equivalent to about 500 g trinitrotoluene (TNT). By contrast, a hydraulic
test of a similar cylinder will give rise only to a small split at a weak point,
with no great release of acoustic or other energy. A compressed liquid, such
as water, contains very little energy by comparison to compressed gas
because of its low compressibility.
Energy storage by means of gas pressure has been employed industrially
8 INDUSTRIAL GASES
Figure 1.2 (a) Gas cylinder after hydraulic burst test. Note typical longitudinal split, (b) Gas
cylinder after pneumatic test. Note gross but basically ductile failure. Few fragments have been
emitted. Photograph from the National Physical Laboratory. Crown copyright 1996.
Reproduced by permission of the Controller of HMSO.
in a number of ways, for example in diesel engine starting or for gyroscope
power in guided missiles. In the future other storage systems might be tried
on a larger scale. Realistic schemes have been devised, for example for using
off-peak electricity during the night, to pump high-pressure air from a
turbo-compressor into underground caverns. During the day the air is
released through a turbine generator. Perhaps surprisingly, such a scheme
might be thoroughly competitive with more familiar schemes involving
water pumping. There is considerable experience with fuel gas storage at up
to 70 bar pressure in underground caverns excavated from salt deposits by
dissolving the salt around a borehole (Melvin, 1988).
1.3.2 Real gases: deviants from the ideal
The ideal gas law is a good guide to the behaviour of gases unless the gases
are at a density comparable to that of the liquid gas or the temperature is
close to liquefaction. The empirical compressibility factor Z can be used in
gas science to allow for deviations from the ideal gas law by introducing it as
follows
Z = 1 for an ideal gas. Tabulations of Z values exist for many gases over a
wide range of pressure and temperature; the result of many years of
repetitive measurements in laboratories around the world. For example, at
around 100 bara pressure and room temperature, the approximate real gas Z
values for air, hydrogen and oxygen are
air 0.984
hydrogen 1.062
oxygen 0.936
The closeness of these values to 1 shows that for the permanent gases at
pressures up to 100 bar the ideal gas equation is still a good approximation.
Rather than relying on extensive tables of empirical data such as Z values, a
number of gas equations based on physical models have been developed. In
addition to giving a deeper understanding of gas behaviour, these are often
convenient in calculation. First, being continuous functions, they can be
used for algebraic manipulations. Second, even where a computer is being
used for calculations, it is convenient to input a few physical model
parameters rather than tediously long tables of Z values. One of the most
useful model-based gas equations of state is that of van der Waals
The ideal gas equation applies to gas molecules that are very small, almost
point-like, elastic spheres which do not interact unless they collide. The van
der Waals equation corrects two of the simplifying assumptions of the ideal
gas equation by allowing the molecules to occupy a finite volume,
represented by the constant b, and by allowing them to be attracted to each
other slightly, the attractive force being represented by the constant a.
Gas mixtures. It is important to be able to predict the behaviour of
mixtures of gases and one of the most powerful concepts in mixtures is
Daltons law of partial pressures. This law follows from the consideration of
the ideal gas equation, for molecules of gases that behave like the elastic
spheres of the ideal gas both when similar and different molecules collide.
10 INDUSTRIAL GASES
Daltons law is that the pressure due to a mixture of gases can be obtained by
adding the partial pressure, of each gas, i.e.
and so on, where is the number of moles of each gas in the volume V.
the pressure exerted by the mixture, is simply the sum of the partial pressures.
Daltons law is subject to similar but more complex corrections than the van
der Waals equation to allow for different forces between different pairs of
molecules. For example, a hydrogen molecule will have a smaller
intermolecular attractive force for another hydrogen molecule than it has for
a fluorine molecule: and explode on contact. For dilute solutions of one
gas in another, however, Daltons law is generally very accurate and it can be
used as a guideline for the behaviour of mixtures of condensible vapours and
liquefiable gases where the ideal gas equation itself is not applicable.
In practice, Daltons law is a valuable principle and guideline but empirical
measurements are essential at higher pressures and where easily liquefied
gases are concerned. In particular, predictions about condensation from
non-dilute mixtures are often erroneous (see, for example, the comment on
the 50/50 nitrous oxide/oxygen medical gas mixture ENTONOX, section
3.9.1).
In addition to Daltons law for pressure, there are other approximate rules
for estimating other properties of gases which can in many cases avoid tedious
experiments by interpolation between the properties of pure gases.
Interpolation is least reliable for mixtures of gases where one or more of the
gases is near to its boiling point or at high pressure, or where the gases are in
roughly equal parts. However, for gases at high temperatures or low
pressures, or for dilute mixtures, interpolation is acceptably accurate.
Interpolation of mixture properties is most accurately carried out for
density thermal conductivity (k), viscosity and thermal capacity
using the following approximate formulae
where etc., are the volume fractions of gases 1, 2 and 3.
These interpolations each apply a slightly different curve of property
calculated versus the molecular proportion of gas 1 in gas 2. The rationale
for the (linear) density calculation follows from Daltons law. The rationale
for the thermal capacity calculation follows from noting that is the
thermal capacity per unit mass and therefore gives the thermal
capacity per unit volume.
For thermal conductivity and viscosity the story is much more complicated
and the expressions above are very rough. Many mixtures of gases, for
example, show a viscosity or thermal conductivity above that of either
component. In these cases it is insufficient to know the parameters of each
gas on its own: parameters related to the interaction of gas 1 with gas 2 are
needed. Accurate formulae are available but involve many of these
gas 1gas 2 interactions as additional parameters. For cases needing good
values of these properties, it is necessary to consult detailed learned
literature for these interaction parameters and/or to obtain experimental
measurements.
1.3.3 The finite size of gas molecules
It is perhaps surprising that the finite size of gas molecules, and their finite
number in a given macroscopic volume of gas, is important in various effects
which are important industrially, e.g.
the transition of flow from fluid to molecular-type flow in small orifices

and in vacuum systems;
the rate of interdiffusion of gas molecules;
the Brownian motion of dust particles in gases, and hence the perform-
ance of filters of very small (> 0.1 particles.
Avogadros number, today defined as the number of atoms in 12 g of carbon
(or 22.41 of an ideal gas at normal temperature and pressure), is
6.022
1.3.4 Heat capacities
The heat capacities of all ideal gases should be identical: energy in an ideal
gas is contained only in the kinetic energy of the molecules and this should
increase with temperature in the same way for all molecules, irrespective of
mass.
However, real imperfect gases differ quite markedly from each other in
heat capacity, and some vary with temperature. It turns out that molecules
which are significantly non-spherical can store energy not just in their
translational motion but also in their rotational and vibrational motion. The
plausibility of this can be enhanced if a gas is envisaged as consisting of N tiny
billiard balls tied by a string to another N neighbours. If the string is very
12 INDUSTRIAL GASES
long (and does not get tangled up) then no effect is noticed, except that the
gas has 2N particles and twice as much kinetic energy. However, as the string
is shortened it begins to constrain the relative motion of the two tethered
molecules and energy is reduced somewhat but energy will go into the
stretching of the string. If the string is made so short that the balls are
overlapping, then the kinetic energy reduces to that of N particles and the
behaviour of a gas of N spherical particles is seen.
In an ideal gas, translational kinetic energy from thermal random motion
will be roughly evenly distributed amongst the molecules and is proportional
to temperature. The Boltzmann constant k times the absolute temperature
(in K) gives the scale of this energy per molecule. The translational kinetic
energy wi l l be 1/2 kT for each molecule or, for an Avogadros number of
molecules(l gramme mole or mole), 1/2 NkT= 1/2 R T, where R is the gas
constant, around 8.31 J
In an ideal monatomic gas the gas will have a specific heat (at constant
pressure, which is normally the most relevant specific heat), of 5/2 R.
The specific heat at constant volume, a less useful quantity, is given by
3/2R. This corresponds to R/2 per degree of freedom i.e., in this case, the
three orthogonal elements of translational kinetic energy at R/2 each plus R
for the energy expended as the gas expands doing work on its containing
walls in the constant pressure situation. For two atoms the gas will also store
energy in rotations in two possible orthogonal axes and the expected is
then 7/2R. On this basis, for polyatomic molecules the addition of another
axis of rotation raises to 8/2R.
Energy stored by compression of the atomic bonds within the molecule
can also give rise to another degree of freedom for each independent mode
of vibration. (For 2N molecules with very loose strings, the result is
obviously 10/2R.) It turns out that quantum mechanics affects the specific
heats of gases by allowing only quantised energy levels in both rotation and
vibration of molecules. The rotation levels are closely spaced so that few
differences are seen between billiard ball behaviour and real atomic
behaviour. However, vibration energy levels are less closely spaced,
perhaps several tenths of an electron volt apart, although the larger the
molecule, the more closely spaced they are. This means that ordinary
thermal levels of energy, around 1/50 eV, can partially excite some of
vibrational modes, depending on temperature, and this leads to higher
values than expected on the basis of the above analysis and the variation of
heat capacity with temperature, especially for large molecules.
To summarise, molar heat capacities at constant pressure of typical
gases at room temperature approximate to
monatomic ideally 5/2 R, e.g. He = 2.50R
diatomic ideally 7/2 R, e.g. = 3.47R
polyatomic ideally 8/2 R, e.g. = 4.11R
These approximations break down for most large polyatomic molecules. For
example, ideally butane would have a molar heat capacity of 8/2R, actually it
is 11.73R. This is because polyatomics such as butane are not rigid bodies but
something between rigid bodies and a cloud of individual particles.
The adiabatic expansion and compression of gases is an important process
in gas technology and one that is governed by the ratio often
designated or g. The specific heat ratio can be calculated from by taking
1.3.5 Bernoulli and viscous gas flow
The designer of typical pipework systems does not need to know all the
intricacies of aerodynamics, although for laminar flow aerodynamic prin-
ciples are actually rather simple, but a basic knowledge of the relation of gas
pressure and flow is very useful.
For many purposes gas flow can be considered to be laminar, i.e. the gas
flows smoothly, without turbulent eddies, and is incompressible, i.e. the
density of the gas varies only slightly because the absolute pressure varies
only slightly. Under such conditions trails of smoke travel in straight lines
which are approximately streamlines (mathematically streamlines are lines
parallel to the instantaneous velocity of the gas).
One of the simplest and most useful relations in gas flow is known as
Bernoullis equation. It applies for incompressible laminar flow and
describes the pressure changes accompanying velocity changes along
streamlines
where P is pressure, is density, is the stream velocity along the streamline
and K is a constant. As a simple example of this equations application,
consider the effect of a smooth constriction in a pipe: to maintain the mass
flow constant the flow will have to speed up through the constriction. It
might naively be expected that the pressure will build up at the constriction.
In fact the pressure will, according to the above equation, drop within the
constriction and then recover back to its original pressure again. This can be
understood if a single element of fluid is followed along its journey through
the constriction: in order to accelerate to a higher stream velocity, there
must be a downward pressure gradient before the constriction. Conversely,
in order to decelerate afterwards, there must be an upward pressure
gradient and hence there must be a lower pressure in the middle of the
constriction. This simple principle is behind the working of the venturi flow
meter.
Bernoullis equation can also be used for pressure drop across an orifice
plate, provided that the orifice is large enough that the pressure drop is small
compared to line pressure (i.e. the flow is effectively incompressible). In the
14 INDUSTRIAL GASES
case of an orifice plate, of course, the pressure dropped is not recovered as
the gas leaves the orifice but is lost in frictional effects as a result of the
turbulence downstream.
1.3.6 Poiseuilles equation
At lower flow rates in pipes viscous flow predominates and the simple
formula of Poiseuille gives pressure drops accurately. Considering the
simple situation of a volume rate of flow Q of gas of viscosity u down a long
circular pipe of constant radius R, with pressure difference along length
L, the following relation applies
Alternatively, in terms of the mean fluid velocity and giving pressure
drop as the dependent
These formulae can both be described as Poiseuilles equation. However,
they should only be applied for the case of viscous flow. To know if flow is
viscous, the Reynolds number (section 1.3.7) should always be calculated.
Motion of dust particles in gases. Another of the simpler formulae in
laminar flow is for the calculation of drag on solid bodies in a gaseous
medium. The Stokes formula is exact for spheres and is commonly used for
the calculation of the motion of small particles of contaminating dust of any
shape in gases
where is visocity, R is the radius of the sphere and is the sphere speed
relative to the gas. A particle will accelerate under gravity until it reaches a
speed at which the drag is equal to its weight, the so-called terminal
velocity. This is given by
or
where is the effective density, i.e. the particle density less the fluid
density, thus correcting for buoyancy.
These formulae will only work, of course, for particles which do not go so
fast as to enter turbulent flow, which begins to have effects past Reynolds
numbers of 0.2. For air, with particles of density 2.65 (e.g. sand), typical
settling velocities are 2.7m per day for 10 particles and proportionally
slower for smaller particles. In air at 1 bar the Reynolds limit is approached
by 200 particles, drifting at 12mm However, the Reynolds limit is
more rapidly approached in denser high pressure gases: 50 particles
would be on this limit at 100 bar.
1.3.7 Turbulent and laminar flow
Laminar flow aerodynamics, although simple in principle, are often complex
in practice. Turbulent flow aerodynamics are complex in principle and
practice. Fluid flow in general involves rotation of fluid elements in vortices.
The simpler systems, such as Poiseuille flow, are irrotational, a simpli-
fication which fails to describe the conspicuous rotation seen in turbulent
flow. The most useful thing that can be done is to learn about the simpler
principles together with some rules of thumb and empirical equations.
Armed with these many pipework problems can be understood and
approximately, if not accurately, predicted.
Liquid flow, like gaseous flow, can also be easily calculated if it is laminar
or viscous but is less tractable if it is turbulent. An easy trap to fall into is to
assume that liquid flow, because liquids have much higher viscosities than
gases, is invariably viscous. Athough liquids do commonly show laminar
flow, cryogenic liquids are frequently exceptional. The low viscosity of
cryogens, all less than 1/5 that of water, means that they often flow with
turbulence under normal conditions.
1.3.8 Compressible flow
With the added dimension of compressibility, fluid dynamics is so complex
that the best that can be done is to enumerate the basic equations for simple
systems and then leave most practical calculations on the complex situations
seen in practice to computer models. A few situations can satisfactorily be
calculated analytically while in addition a few useful rules of thumb, albeit of
limited applicability, exist.
One useful rule of thumb, for example, concerns the flow of gas through a
narrow orifice. Consider an orifice restricting the flow of gas to the
atmosphere (this situation pertains to the real situation for some gas
cylinders which are fitted with an orifice to limit the maximum flow rate). As
the flow tends to increase to near sonic speeds according to the Bernoulli
equation, the flow through an orifice becomes choked or critical, i.e. a
decrease in the downstream pressure will have no effect on the flow because
the influence of the downstream pressure can only propagate at the speed of
sound and therefore cannot pass back through the orifice. In this simple
situation the flow rate can be expressed as follows
16 INDUSTRIAL GASES
where C is the velocity of sound,
Hence
Where P, and are pressure, density and specific heat ratio of the
gas in the reservoir A is the orifice area and k is a constant.
2
That the
formula should involve the ratio (constant pressure/constant volume
specific heats) should be no surprise as rapid expansion across an orifice is
clearly adiabatic.
Compare the above with the non-compressible equivalent formula for an
orifice
Here the downstream pressure P is important as the flow is subsonic and
downstream events can affect upstream flow rates. k is a constant related to
the ratio orifice area/vena contracta area, which is roughly constant. Vena
contracta is a reference to the shrinking of the cross-sectional area of a
stream issuing from an orifice.
Pipe flow with a high pressure drop is also sometimes calculable by
formula. This may be needed, for example, when using very small capillary
sample pipes or at low pressure. The relatively simple formula for this is
where k is a constant depending on the gas and its temperature, and L is the
length of the pipe. (Compare this with Poiseuilles equation
If mass flow rate is desired, then density, of which an average is
must multiply the formula and Poiseuille then gives
where is another constant.)
Dimensionless groups and Reynolds number. Finally, before leaving
simple numerical formulae and admitting that further problems require
numerical solution with a computer, the possibilities of a dimensional
analysis of a gas flow problem should be examined. Dimensional analysis,
even where it does not give answers, can restrict the amount of experimen-
tation which may be necessary and suggest ways in which a problem can be
scaled down to suit laboratory experiment. Dimensional analysis may well
2
A more exact formula replaces with
provide the required cross-check on the realism of computer-generated
results.
Dimensionless groups have proved useful in much of physics but
especially so in fluid dynamic problems. The most commonly used
dimensionless group is the Reynolds number. This expresses the balance of
viscous and dynamic force on elements of a gas stream. A high Reynolds
number means that dynamic forces, the mass inertia forces of the flow,
exceed the viscous forces within the gas. The result will generally be
turbulent flow. A low Reynolds number means that the viscous forces far
exceed inertia effects and the resultant flow pattern will be indistinguishable
from that of treacle. The Reynolds number, Re, is defined as
i.e.
where is the density, V is the mean velocity, L is the characteristic length
and u is the viscosity.
For pipework Re <<2000 will lead to laminar flow whilst for Re >>2000
turbulent flow will occur. However, the boundary between the flow regimes
is not clear-cut and flow in the whole region around may be
difficult to predict. If a laboratory experiment is to be used to model
full-sized equipment, and flow will occur near the turbulent/laminar flow
boundary, then by designing an experiment where the Reynolds number will
be similar to that at full-scale realistic results should be obtained.
Design of a linear-flow element demonstrates a simple application of the
Reynolds number. Suppose the flow in a pipe is V per second. If the
Reynolds number is 10 000, turbulent flow would be expected, giving a
characteristic pressure drop depending on the square of the flow rate, If
that single piece of tube is replaced by 10 000 small diameter tubes adding up
to the same cross-sectional area, then, at the same flow rate Q, the mean
velocity will be the same but the pressure drop will vary linearly with the flow
rate Q, because the Reynolds number will be only 100, with viscous flow
predominating.
Linear-flow elements are designed to impose a linear-flow versus pressure
curve on a gas flow by ensuring viscous flow. They do this by the use of a
multiplicity of small gas passages. Reynolds number calculation reveals why
this tactic works. Basically, by restricting the flow in this way, the effects of
dynamic inertia are made small compared to the large viscous drag required
to force the gas through the tiny passages.
Dimensionless groups in general make comparisons between the forces
on an element of gas e.g. inertia, viscosity, gravity, pressure, etc. Other
common dimensionless groups are:
18 INDUSTRIAL GASES
Nusselt number, hD/k: heat input/heat lost from conduction, relating to

heat transfer
Prandtl number, Cu/k: kinematic viscosity/thermal diffusivity, i.e. the

ratio of diffusion of momentum to diffusion of heat, relating to forced
convection (contains only fluid properties and is substantially constant
in many gases under normal conditions);
Grashof number, buoyancy force/viscous force, relating to

natural convection;
where h = coefficient of heat transfer, D = diameter, k = thermal
conductivity, C = specific heat, u = viscosity, g = gravity, b = thermal
expansion coefficient, = temperature difference and = density.
Turbulent flow in pipework. With large Reynolds numbers pipework can
display very predictable turbulent flow. This situation is described by
another useful rule of thumb by Darcy. Darcys (or Fannings) rule is for
well-developed turbulent flow in pipework and is an approximate rule
analogous to Poiseuilles accurate equation
where f is the empirical friction factor, v is the mean velocity, is the
medium density and L is the length of pipe. At low Reynolds numbers the
factor f is equal to 8/Re, which will give Poiseuilles equation as before.
For a range of Reynolds numbers from 2000 to 100 000 and smooth pipes
For rougher pipes and higher Reynolds number, the friction factor f will
depend on the roughness ratio k/2r, where k is the roughness size, but is
independent of the Reynolds number. Values are typically
0.04 0.016
0.01 0.01
0.04 0.07
0.001 0.005
0.0004 0.004
0.0001 0.003
If the value of f predicted by the roughness ratio is less than that of the
smooth pipe formula, then the smooth pipe formula is used.
It is interesting to note that with the f factor constant, resistance to
turbulent flow is independent of the fluid viscosity. Even a superfluid, a
fluid of zero viscosity, thus shows resistance to flow at high flow rates.
Cryogenic liquid gases, such as liquid nitrogen, as noted in more detail in
section 3.7.2, have low viscosity whilst liquid helium can actually be
considered as a mixture of superfluid and a normal fluid of low viscosity. In
k/2r f
practice both of these frequently show large resistance to flow, contrary to
what might be expected from their viscosity, because at fairly ordinary
pumping speeds their flow is turbulent.
1.3.9 Computational fluid dynamics
Computer numerical models, such as the popular FLUENT and PHOENIX
codes, are now generally available in industry and there are versions that can
be run on desktop workstation personal computers. These simulate gas
behaviour by setting up finite elements of a gas and applying fundamental
equations to the motion and state of these elements, with boundaries
defined by the geometry of the parts contacting the gas. They are capable of
simulating flow problems, including turbulence, compressibility and ther-
mal effects, and will even allow calculations for reacting gases, such as in
combustion. Most of these situations are impossible to calculate analyti-
cally. Computer models allow fairly accurate calculation of the performance
of components of geometry which would not be possible using analytic
methods, even for simple conditions such as laminar flow.
The simplest computer models for steady laminar flow simply provide
solutions to Poissons equation, with a few boundary conditions (e.g. V = 0
for y < 0 and V = 100 for y > 10).
where lines of constant V are streamlines and f ( x) is simply zero in the
simplest case. (In the past, steady laminar flow streamlines have been
plotted using electric fields across conductive paper, a kind of analog
computer apparatus. This is possible because electric fields also obey
Poissons equation. The boundary conditions are applied by means of a
battery and pieces of metal foil.)
Even the more complex computer models still have a relatively simple
basis, such as the general partial differential equation
where is the density of the medium, t is time, V is the velocity vector, is
an exchange coefficient, is a source term (e.g. pressure gradients
and if is a velocity component) and is the dependent
variable to be computed (e.g. the velocity component gives the
equation the continuity equation, which simply says that
matter is neither created nor destroyed in the flow pattern).
More complex equations are needed for simulating multiple-phase flows.
20 INDUSTRIAL GASES
The most sophisticated computer models can solve problems with other
complications such as chemical reaction, thermal radiation, transonic or
supersonic flow and porous media.
Like the simple Poisson equation these equations are solved by first
deciding on boundary conditions, i.e. where the edge of the gas contain-
ment is, velocities, densities or pressures of gas sources, etc. Then gas
equations of state must be entered into the model. In many cases a simple
formulaic equation of state, such as the van der Waals model, may be
applied. The differential equations must then be turned into algebraic
equations which can be solved numerically with computer programs.
Such differential equations are applied to each finite volume element of
a three-dimensional grid, chosen to be small compared to the scale of the
phenomenon studied, and applied to successive time intervals, chosen to
be short compared to the characteristic time of the phenomenon. The
computer must simulate all the forces acting on the volume element
(pressure on each face, turning moments) and calculate the reaction of the
element (acceleration, change of temperature and volume). The influence
of that element on its neighbours must then be calculated. Finally, the
results must be displayed in a meaningful way to the operator to show flow
behaviour. This is illustrated in Figure 1.3, which shows an example of a
display from a CFD calculation of turbulence at a T-junctions in a gas
pipeline.
The different computer programs, such as PHOENICS and FLUENT, differ in
the detail of their programming and in how they approach the transform-
ation from differential to algebraic equations. PHOENICS, for example, uses
the finite-volume method developed in Los Alamos, USA, and Imperial
College, London. It straightforwardly applies a grid and integrates each
differential equation for each volume element to give an algebraic equation
which is then solved.
The workload on the computer is considerable even with small grids and
few time slices, and quickly grows as the grid is divided more finely. Many
computational tricks have to be used in order to use computational fluid
dynamics (CFD) on desktop computers. Desktop computers are sufficient
for many problems but it should be noted that quite large and expensive
parallel-processor computers are needed for problems that are apparently
only fractionally more difficult than those solved on the desktop computer.
There is a considerable amount of skill in applying CFD successfully and
economically. For example, by choosing a nonlinear grid or two dimen-
sions instead of three, computation time can be greatly reduced. In
problems with rotational symmetry where turbulence is absent or un-
important, for example, a cylindrical grid, using only two dimensions, will
be possible. In problems of steady-state flow, the first transient term can be
omitted.
The use of computer modelling of flow is now almost mandatory for any
Offset T-junction scalar equation
Velocity vectors on the mid-plane of a T-junction
Figure 1.3 Example of CFD calculation of turbulence at a T-junction in a gas pipeline (courtesy
of CHAM, London).
gas apparatus that is at all critical. However, when using computer-
generated results a cross-check of some kind, from experiment, from
simplified theory or perhaps even from a different computer model, may
well be useful. Wherever possible, at least the order of magnitude of a
reported quantity should be cross-checked.
22 INDUSTRIAL GASES
1.3.10 Diffusion
According to kinetic theory the diffusion constant D of an ideal gas is given
by
where L is the mean free path (see section 1.3.11) and is the mean
molecular speed. Substituting standard formulae for L and the diffusion
constant is therefore given by
where n is the molecular number density, per unit volume, k is Boltzmans
constant, T is absolute temperature, M is molecular mass and S is the
molecular cross-sectional area. The diffusion constant is thus inversely
proportional to the pressure (the number density being proportional to
pressure at ordinary pressures), i.e.
Many time-varying diffusion problems may be solved by application of the
diffusion equation, which in one dimension is
This is derived from the steady-state Ficks law, which defines the diffusion
constant D as where J is the flux of molecules diffusing and
is the concentration gradient.
The behaviour of mixtures in gas cylinders is a good example of how
diffusion can achieve mixing over time. Suppose a cylinder of gas is filled with
two gases more or less perfectly separated at time zero. The lighter gas is at the
top, the heavier at the bottom. The mixture will slowly form at the interface
and the progress of the mixed band through the cylinder will spread to a width
proportional to the square root of the time passed. This curious time
behaviour is a consequence of the mathematics of the diffusion equation.
The diffusion equation frequently has solutions of the form
where a is a constant. These solutions show the characteristic diffusion
equation property that the width over which a gas has diffused increases as
the square root of time.
A typical 5 m
3
200 bar gas cylinder will be fully mixed for most purposes
after a few months. However, inspection of the above equation will reveal a
simple way to reduce the mixing time dramatically: simply lie the cylinder on
its side. The gases will then take up the same positions with the lighter gas on
top of the heavier gas. However, in this case the distance over which
diffusion must take place is much shorter, a factor of around 10 less. With a
characteristic distance 1/10, diffusion will take place in or 1/100 of the
time, a day or so for a typical cylinder.
The calculation of diffusion in textbooks often gives an example of
diffusion at atmospheric pressure, which may give an illusory idea of the
rapidity of diffusion in all situations. Diffusion, because it is inversely
proportional to pressure, is much slower at the higher pressures often used
in industry. In a gas cylinder at 1 bara, for example, the textbook example
will give 2 m diffusion distance in 1 h. However, with full cylinder at 200 bar,
the diffusion distance for 1 h is only 14cm and, furthermore, this distance
only widens as the square root of time, so 200 h is needed for the equivalent
diffusion state at 200 bar. Figure 1.4 shows an example of diffusional mixing
in a gas cylinder.
It is sometimes asked whether a gas mixture, particularly a light/heavy
mixture like sulphur hexafluoride in hydrogen, once formed will unmix: the
answer is basically that it will not. The science of statistical mechanics, which
explains why this is so, is well-described in standard texts, although popular
books such as Ruelle (1991) may deepen understanding. Provided the
constituents are not near their boiling point, and cannot therefore condense,
a gas mixture will remain mixed. In fact, many mixtures with condensible
components below their normal boiling point will not unmix either because
the partial pressure of the condensible component is below 1 bar, the
reference pressure for the boiling point.
There is a certain amount of gravitational separation in gas mixtures but in
normal earth gravity this is entirely negligible
where P
i
is the partial pressure of the ith gas, g is the acceleration due to
gravity, h is the height above sea level (at which k is Boltzmanns
constant and T is the absolute temperature.
e.g.
at the bottom of the cylinder, P
1
/P
2
= 1.001 at the top.
In high-speed centrifuges this is no longer true and today the most efficient
method yet discovered for the separation of uranium isotopes is gravitatio-
nal separation of UF
6
using centrifuges.
1.3.11 Molecular flow: small orifices and vacuum systems
There are now many industrial processes employing gases at low pressure.
Many are in the processing of microelectronic circuits, although metal/
plastic coatings, glass coatings, engineering chemical vapour deposition
(CVD) coatings, flatscreen displays and solar cells are also users of low
pressures. Books on vacuum techniques are recommended (e.g. Dennis and
Heppell, 1968) for a detailed account of the behaviour of gases at
high-vacuum pressures.
24 INDUSTRIAL GASES
Figure 1.4 Diffusional mixing in a gas cylinder.
However, many processes operate in the no-mans-land between
pressures where true vacuum technique applies and pressures where gases
behave more or less as they do at atmospheric pressure. An example of this
is the evacuation of gas cylinders before filling them (for high purity grades).
In this pressure region it is useful to know about vacuum as well as
atmospheric pressure characteristics.
Gases in ordinary size pipes and vessels (an inch or so) switch from more
or less familiar viscous or turbulent flow to the radically different regime of
molecular flow at around 0.1 millibar. Below this pressure gas molecules
may travel from one side of the pipe to the other without colliding, giving
radically different flow characteristics. The crucial parameter is the mean
free path i.e. the average distance gone before a molecule meets another
molecule.
The mean free path in air at atmospheric pressure is just 0.05 m.
However, the mean free path is inversely proportional to pressure and at a
typical industrial vacuum pressure, say 1/10 000 of an atmosphere, the mean
free path is about 5 mm.
The mean free path is given approximately by
where n is the number of molecules per m
3
(proportional to pressure at lower
pressures) and is the collision cross-section of the molecule, approximately
where R is the molecular radius.
For a long pipe with steady viscous flow at normal pressures, the mass flow
rate is given by the Poiseuille equation, as above
The molecular flow equation for low-pressure gases is
Figure 1.5 Flow through a pipe showing transition from molecular to fluidic flow.
The molecular flow regime involves a much lower resistance to flow than the
viscous calculation gives. As the pressure sinks very low, the molecules can
travel without collision along a long length of pipe, so this phenomenon is
perhaps unsurprising. However, the lower power in pipe radius means that
moving to a larger pipe does not offer as much advantage as in viscous flow.
The absence of an average pressure term reflects the fact that the net flow,
Q, is simply the difference between the molecules leaving the high-pressure
end and those leaving the low-pressure end.
The evacuation of a cylinder to millitorr (10
6
bar) pressures involves a
transition from molecular to fluidic flow (Figure 1.5). If an approximately
constant volume flow rate from the vacuum pump is assumed, then, in the
absence of resistance to flow from pipes, the cylinder will follow an
exponential fall in pressure
where S is the pumping speed in V is the cylinder volume in litres and is
the time in s. However, the effective pumping speed, S, is affected by the
resistance to flow, which is usually expressed in vacuum problems in terms of
conductance. The conductance of a pipe, C, in gives the flow with a
differential pressure (traditionally given in torr) as follows
Q is thus the volume flow in Torr It is convenient to define C in this way
because then the effective capacity of a pump connected via a pipe of
conductance C can be simply calculated as
If the pressure is higher than around Torr, the conductance, C, is not
constant, in fact it increases exponentially. This results in a steeper initial
26 INDUSTRIAL GASES
pressure drop when is high, followed by a slower exponential decrease
reflecting the fact that is small, being dominated by the large term C. (In
practice a curve of this sort is distorted below a certain pressure by the
outgassing of water vapour absorbed onto the container and pipework walls.
In our example, to achieve high vacuum the cylinder must be baked to 100C
or more.)
1.3.12 Heat transfer in gases
Heat transfer is a substantial branch of chemical engineering in its own right
and standard texts such as Fraas and Ozisik (1965) and the usually
substantial sections in books on chemical engineering, such as Perry (1984),
are suggested for a deeper understanding. However, there are a number of
simple situations which frequently arise in the use of gases where straight-
forward calculations are useful. Many of these involve the transfer of heat
from solid surfaces to the gas and vice versa. Heat transfer in gases takes
place via three mechanisms: radiation, convection and conduction. Before
starting the detailed solution of any problem, the relative importance of the
three mechanisms should be roughly estimated
Radiation typically depends more on the nature of the solid surfaces than
on the gas between them. It is usually unimportant for shiny polished
surfaces, especially at low temperatures.
Whether convective heat transfer can take place depends on whether

effective circulation of the gas can occur, either forced with a fan or via
natural convective flow. In a geometry with a hot surface above a cold
one, natural convective heat transfer will be very low.
Conductive heat transfer is the most readily calculated of the three

mechanisms. It depends only on the geometry of the surfaces and their
temperatures, and the nature of the gas. Only in high vacuum does
conductive transfer reduce to zero, although in many problems it may not
be as important as the other two mechanisms.
1.3.13 Thermal conductivity and viscosity of gases
For normal pressures the thermal conductivity of a gas, like the viscosity, is
independent of pressure. This arises as follows. The thermal conductivity,
K, of a substance is the constant K in Searls equation for one-dimensional
heat flow along a bar
where dQ/dt is the heat transferred along the x direction. The rate at which
heat (molecular kinetic energy) is carried along is clearly proportional to the
heat capacity, of each molecule, the molecular velocity, and the
number of molecules per unit volume, n. Less obvious perhaps is the effect
of the mean free path to which the thermal conductivity is also proportional.
The mean free path, L, is proportional to 1/(nS), where n is the number of
molecules per unit volume and S is the molecular cross-section.
Hence
In fact, the accurate formula for ideal gases is
where K is the conductivity, is the specific heat per mole, is the mean
molecular speed, N is Avogadros number and S is the molecular collision
cross-section, i.e. the area around a molecules centre where a collision will
occur. The only variable in the equation is the velocity, which is simply
related to the absolute temperature by
i.e.
which is manifestly independent of pressure. In fact, between 1 mbar and
about 100 bar, a range of pressures from vacuum to the highest normal gas
pressure, the conductivity of a gas, like the viscosity, is remarkably constant.
This has important implications for thermal insulation: a Dewar vacuum
insulated flask, for example, must employ a high enough vacuum that this
constancy of the thermal conductivity breaks down. The thermal conduc-
tivity of a gas decreases with molecular weight and size (Table 1.1).
28 INDUSTRIAL GASES
Kinetic theory for the viscosity of gases gives the result
where u is the viscosity, m is the molecular mass, is the mean molecular
speed, and S is the collision cross-section (the area within which molecular
vectors are collision paths). The only variable in the equation is the velocity,
which is simply related to the absolute temperature by
Hence
Kinetic theory thus predicts that the viscosity of an ideal gas does not vary
with pressure. In fact, for real gases at ordinary pressures, the viscosity does
indeed depend very little on pressure. Between 1 mbar and about 100 bar, a
range of pressures from vacuum to the highest normal gas pressure, the
viscosity of most gases is remarkably constant. Even though the gas at 100
bar has 100 000 times as many molecules per as it has at 1 mbar, the
viscosity, and hence the frictional resistance to motion, is the same. An
additional factor that must be taken into account, however, is the greater
likelihood of turbulence at higher pressures because the Reynolds number is
proportional to density and hence pressure.
The formula above also predicts that gas viscosity should vary with
This is inaccurate and a correction to this formula (from Sutherland, see
Kaye and Laby (1995)) gives
where B and C are constants.
Small heavy molecules are expected to have higher viscosities than large
light ones and this is borne out in practice, with the noble gases all showing
high viscosities and larger hydrocarbons low viscosities (Table 1.2).
The high viscosity of the monatomic gases has an impact on the use of
gases as thermal insulators. In tiny cells, such as in plastic foam, the best
thermal insulating gas will be that with the lowest thermal conductivity,
normally a heavy polyatomic gas. However, in larger cells convection will
take place, with the less viscous gases, such as the heavy polyatomics,
convecting more and providing a poorer insulation than those of high
viscosity.
1.3.14 Vapour pressure of liquids
Many gases are used in liquid form either under pressure or at low
temperature. The liquid (or solid) form of a substance, if kept in a container
with an ullage space, will eventually reach a dynamic equilibrium with the
gaseous form of that substance. This equilibrium pressure, the saturated
vapour pressure, is a straightforward function of temperature only.
Normally when a liquid is heated gently molecules simply leave the liquid
surface for the gas phase and evaporation is said to take place. If a liquid is
heated more vigorously bubbles form in it when the saturated vapour
pressure of the liquid exceeds the gas pressure and the liquid is said to boil.
The converse process of condensation of the gas occurs during cooling and
can also take place either in the body of the gas or at a surface.
In mixtures the temperature at which condensation takes place (the dew
point) is cooler, often by several degrees, than the temperature at which
bubbles form (the bubble point). For example, the bubble point of liquid air
is 78.75 K, whilst the dew point is 81.75 K. Bubbles forming at the bubble
point are rich in the lighter component whilst the liquid condensing at the
dew point is rich in the heavier component of the mixture. (In the case of air
condensing on a liquid nitrogen line, for example, the liquid forming on the
outside of the line will be rich in oxygen, a phenomenon that can give rise to
accidents if oxygen-rich liquid air drips on to organic material below.)
The boiling or condensing of a gas is a complex phenomena, with clean
liquids and gases frequently persisting in states far from equilibrium. Pure
liquids may be heated far above their boiling point, or superheated. When
boiling finally occurs in a superheated liquid it is sudden and violent. Pure
gases may be cooled far below their boiling point, when condensation may
be also be very sudden and the vapour is said to be supersaturated.
Both supersaturated vapours and superheated liquids have been used for
the detection and tracking of nuclear particles: the minute amount of
ionisation caused by the passage of a high-energy charged nucleon is
sufficient to act as a series of nuclei for condensation or boiling. If conditions
are correct, nuclear particle tracks are revealed by rows of droplets in the
30 INDUSTRIAL GASES
Figure 1.6 Liquid hydrogen bubble chamber (BEBC at Cern, Geneva) showing boiling around
the expander piston (centre) and charged fundamental particle trails of microscopic bubbles,
curved due to the large applied magnetic field. Sometimes neon was mixed with the LH to
increase the cross-sections for the particle collisions (courtesy of High Energy Physics Group,
Imperial College, London).
vapour in the case of the cloud chamber. The early development (1912) by
Wilson (see Enge, 1966) of the cloud chamber enabled many important
nuclear physics discoveries. In the case of the bubble chamber the tracks
are revealed by rows of bubbles in the liquid, typically liquid hydrogen. The
hydrogen bubble chamber was the most important particle physics analyser
from the 1950s to the late 1970s (Enge, 1966; Perkins, 1972) (Figure 1.6). It
enabled the foundations of the quark theory of matter to be laid.
Supersaturated vapours and superheated liquids have important industrial
applications, too (see section 3.7.2 for further discussion of boiling).
The saturated vapour pressure of many substances can be approximated
to an exponential function of the form
3
This form can be understood in a simple way as follows. There is a spread of molecular speeds
in a liquid, with only a few molecules having enough speed to escape the attractive forces of the
liquid. The fraction with enough speed depends on temperature. The velocity distribution
follows a normal distribution pattern to most statistical distributions and is given by an
exponential of the form
Taking the fraction with energy greater than the escape energy, gives the
exponential pressure law above.
Figure 1.7 Vapour pressures of different substances.
Figure 1.8 Vapour pressure of water.
or
where C and are constants peculiar to the substance, and k is Boltzmanns
constant. In Figure 1.7 (pressure) is plotted against a constant/T. This
plot should be a straight line if this law is followed. For most substances, the
exponential in 1/T is a remarkably good approximation. The temperatures
vary in Figure 1.7 from the intense furnace heat of the 2620 K boiling point of
aluminium to the ultracold behaviour of helium near absolute zero but
32 INDUSTRIAL GASES
deviations from the law are not great. Helium is a very exceptional
substance, however, and is clearly following a slightly different law.
Figure 1.7 shows the vapour pressure of water from 0 to 20C and also
demonstrates the exactness of fit of the exponential function.
1.3.15 Sublimation and supercritical fluids
Most everyday substances are solid at low temperature, melt when heated
and vaporise when heated more strongly. Referring to Figure 1.8, a PT or
phase diagram, most substances follow a line such as BB.
Amongst common industrial gases this standard behaviour is also
followed, with a few exceptions. Solid carbon dioxide (dry ice) sublimes,
going directly from solid to gas, as does acetylene. They have no liquid phase
at atmospheric pressure, following a line such as A A on the phase diagram,
below the triple point. At pressures higher than the triple point, liquid
phases will form for these gases, however, so that, for example, carbon
dioxide at 10 bar or more forms a liquid phase, a fact which is used to
advantage when storing carbon dioxide in bulk (section 3.4.2).
Carbon dioxide is the most common gas which sublimes at atmospheric
pressure. Solid acetylene is regarded as dangerous, being subject to
detonation, so is not normally utilised as a solid but it sublimes at 189 K. A
slightly more unusual gas, sulphur hexafluoride, sublimes at 209 K, while
there are other more reactive and unusual gases which also sublime. It
should be noted, however, that the triple points of gases such as nitrous
oxide, neon and even argon are close to atmospheric pressure and modest
pressure reductions can result in them solidifying, perhaps blocking up
apparatus unexpectedly (Table 1.3).
At high temperatures substances follow a line such as CC in the phase
diagram (Figure 1.9), passing from solid into a state which is dense but has
no surface tension, i.e. a supercritical fluid. Above the critical temperature
Figure 1.9 Schematic P/T diagram for argon.
no liquid can exist, but instead the substance forms a supercritical fluid or
dense gas (Table 1.4).
As liquids are heated and the average energies of their constitutent
molecules grow larger, their surface tension, heat of vaporisation and
viscosity grow smaller, and their compressibility increases. This is intuitively
reasonable on an atomic model of a liquid, in which the higher thermal
energies force an increase in molecular separation and a resultant decrease
in intermolecular forces.
As the critical point is approached, the surface tension approaches zero,
as does the heat of vaporisation, while the viscosity rapidly modulates
downwards by three or more orders of magnitude from liquid to gaseous
values. The compressibility also changes rapidly near the critical point,
increasing to very large values just below the critical point but falling again to
the larger values typical of gases well above (e.g. Reid, 1977).
Supercritical fluids have some important properties. They are dense
enough to hold appreciable amounts of other substances in solution and can
be used as solvents in which surface tension is entirely absent. This enables
them, for example, to be extracted from porous material (carrying dissolved
substances away) without causing the collapse of pores. For this reason
methane, methanol vapour, steam and carbon dioxide, inter alia, are
important industrial solvents under supercritical conditions.
Carbon dioxide has a critical temperature conveniently near room
temperature, 30C. It was the first substance whose near-critical properties
were investigated and can be used as the basis of a demonstration. A suitably
small amount of liquid CO
2
partially filling a sealed strong (critical pressures
are often in the 3070 bara region, is 74 bara) glass vessel (behind a
safety screen) can simply be warmed through its critical point with a gentle
34 INDUSTRIAL GASES
source of heat such as a hair-drier. Under the correct conditions the liquid
meniscus disappears, while a misty appearance (critical opalescence) can
be seen: the consequence of rapid spatial variations in density, which should
be expected as the compressibility of the fluid becomes very large because
micro pressure changes are able to cause macro density changes.
Supercritical fluid extraction. The high compressibility of supercritical
fluids means that the supercritical fluid density can be modulated hugely by
relatively small changes in applied pressure. It is this that allows efficient
continuous extraction processes, supercritical fluid extraction (SFE), to be
designed (Kirk-Othmer (1982) and Irani and Funk (1977) both describe the
process well). A small increase in pressure is used to raise density to obtain
solvent action, while precipitation is carried out by a small reduction in
pressure, when the low density achieved causes solubilities to fall dramati-
cally.
is one of the most popular substances for use in supercritical fluid
extraction because of its combination of convenient properties good
solvent, supercritical at normal temperatures and reasonable (100 bar or so)
pressures, low toxicity as well as its low cost. It is used on an industrial scale
for the preparation of decaffeinated coffee. The undesired stimulant
compound, caffeine, is extracted from ground beans by SFE by steeping to
dissolve the caffeine in high-density supercritical followed by the release
of the caffeine to waste and recovery of by expansion to low density.
1.3.16 Dissolved gases in liquids
In many situations, the very small amount of gas that typically dissolves in a
liquid is not important. However, there are situations where the dissolution
of gases in liquids is a matter of great consequence in nature and of
considerable importance in industry. The small amount of oxygen that will
dissolve in water is vital for the life of aquatic creatures, while the similarly
small amount of nitrogen that can dissolve in human blood is a serious
difficulty for divers. Historically, the different solubilities of oxygen and
nitrogen in water enabled crude oxygen concentrators to be constructed.
More recently, gases dissolved in hydrocarbons became very important as
the oil industry matured and the very large volumes of gases held in most
crude oils, including methane, ethane, ethylene, propane and butanes, are
now extracted and put to useful purposes. In industry many separations or
purifications of gases rely on the absorbtion of one gas from the mixture by
means of a spray of liquid or the trickling of liquid over a packed tower.
Gases dissolve in liquids to a greater or lesser extent, depending on their
chemical compatibility. For example, nitrogen, an inert, nonpolar mol-
ecule, dissolves only very slightly in water. By contrast carbon dioxide is a
polar molecule which is attracted to water molecules and in fact reacts to a
small extent with water to form carbonic acid. As a result, carbon dioxide is
much more soluble than nitrogen in water: at 15C about 1 volume of
dissolves in 1 volume of water, as opposed to 0.023 volumes of N
2
.
The theory for the dissolution of gases in liquids is fairly simple. Many,
particularly the more sparingly soluble gases, follow Henrys law up to 100
or 200 bar. Henry's law states that the pressure or partial pressure of a gas in
equilibrium (P) with a liquid is proportional to the molar concentration (X)
of that gas, i.e.
Figure 1.10 shows the curves for common gases in water, demonstrating the
validity of Henrys law.
The constant, k, in Henrys law varies with temperature and that
variation can frequently be modelled by an exponential factor of the form
where H is related approximately to the energy required to extract a mole of
gas from the solvent.
Figure 1.10 Solubility of common gases in water.
36 INDUSTRIAL GASES
Figure 1. 11 Oxy-haemoglobin saturation curve.
A more limited situation, but one which may be useful, is that of the
vapour pressure of mixed liquids. These often adhere to the more stringent
Raoults law, which gives the constant of proportionality as the vapour
pressure of the pure gas. This applies only in a few cases where the forces of
attraction and repulsion between the two types of constituent molecule are
closely similar, for example mixtures of atomic isotopes such as and
(heavy water) or neighbouring members of an organic homologous
series such as butanol and propanol.
For nitrogen in water and similar cases, the forces of attraction between
the gas molecule and the liquid molecules are weak. Where the forces are
larger and a weak chemical bond can be considered to have been formed, the
straight line of Henrys law response is replaced by a curve with a plateau,
clearly indicating the formation of a more or less stoichiometric chemical
compound. A typical curve shows the amount absorbed steadily rising with
gas pressure until the amount added corresponds to saturation of the
absorber compound. After that point the amount absorbed increases only
slowly with gas pressure. The solubility of oxygen in blood is an example of
this (Figure 1.11).
A lot of experimental work has been done on the solubility of gases in
liquids. Much data are of poor quality, however, observations having been
made with apparatus of widely varying construction and effectiveness, and
the data are often quoted in awkward units. A good account is given in Fogg
and Gerrard (1991), which tabulates and graphs a lot of valuable data.
Amongst this data the high solubility of carbon dioxide in oxygenated
hydrocarbon solvents is striking, as is the solubility of oxygen in fluorocar-
bons.
Finally, it is worth noting that gases dissolve in solids as well as liquids.
In the case of solids, however, it is much more difficult to reach the
equilibrium solution because of the slow nature of the diffusion of gases
through solids. The case of most interest is that of thin plastic films, where
gases dissolve in the film and then diffuse through (see sections 2.1.7,
2.2.11 and 3.8.12).
1.4 Some specific gases
Countless organic compounds are possible and millions of inorganics.
However, there are only a few hundred compounds which exert a vapour
pressure of more than 1 bar at 20C, and qualify for the name gas.
Furthermore, few of these compounds have importance in industry. A brief
introduction to a few of them follows.
Nitrogen
Molecule: formula NN
weight 28.01
% in air 78
Physical: boils 77 K
melts 63 K
critical temperature 126 K
Liquid properties: density
viscosity
tonnage cost $15
heat of evaporation
Reactions: mainly inert
500 bar catalyst
iron nitrides at 1000C
Mg burns
Discovered by Daniel Rutherford in 1772, nitrogen is the principal gas in air
(78%). Its first major industrial use lay in the manufacture of ammonia, used
within the chemical industry as an intermediate for many compounds but
especially nitrogenous fertilisers.
Oxygen
Molecule: formula OO
weight 32.0
% in air 21
38 INDUSTRIAL GASES
melts 54 K
viscosity
tonnage cost $15
heat of evaporation
Reactions: burns/oxidises most elements and
all organic compounds
hottest flame
iron burns
alcohols forms aldehydes/acids
Discovered in the mid 1700s, oxygen is the most important constituent gas of
air (21.6%). Its first major industrial use was for limelight, intensifying gas
flames until they could be used to heat calcium oxide (lime) to white heat.
Liquid oxygen is the only common liquefied gas to be coloured; it is pale
blue.
Argon
Molecule: formula Ar
weight 40
% in air 1
melts 84 K
viscosity
tonnage cost $500
Reactions: inert
Discovered at the end of the Victorian era, argon only became important
with the rise of highly controlled smelting processes and with the growth in
demand for high quality arc welding. Arc welding with powder-coated
sticks, the powder forming a fluid slag or flux or evolving CO
2
or similar
gases to shield the molten weld pool from oxidation, had been devised
around in 1911 or 1912. However, the quality in these processes was not
particularly good and patents were taken out in the 1930s for argon/
hydrogen and similar inert gas mixtures. The full development of the
potential for argon for welding only occurred after the Second World War,
however, with the development of new processes using other gases added to
argon and the arrival of inexpensive argon in large quantities.
Carbon dioxide
Molecule: formula NN
weight 28.13
% in air 300 ppm
Physical: sublimes 195 K
Liquid properties density
(at 10 bar): viscosity
tonnage cost $15
Reactions: mainly inert
500 bar catalyst produces CO
Iron 1000C mildly oxidising
Discovered in the mid-1700s, carbon dioxide has a high heat capacity and
sublimes directly from solid to gas at atmospheric pressure. It was therefore
used early in solid form (dry ice) as a refrigerant.
Acetylene
Molecule: formula CH=CH
weight 26
Physical: sublimes 189 K
viscosity
tonnage cost $1500
(liquid acetylene is a dangerous,
explosive, liquid)
Reactions: highly exothermic when it reacts
oyxgen gives hottest flame
many active chemicals have
addition reactions, e.g. with HC1
to give vinyl chloride
Discovered as the result of adding water to calcium carbide chips in the
1800s, the gas is still mostly prepared in that way today, although
petroleum-derived acetylene is important, too.
Methane
Molecule: formula
weight 16
40 INDUSTRIAL GASES
Physical boils 112K
melts 91K
viscosity
tonnage cost $15
Reactions: mainly inert except for combustion
Methane was recognised as the flammable principle of marsh gas early on
and as a constituent of natural gas and petroleum in the 1850s but was
treated largely as an undesirable by-product of petroleum until the 1950s.
Many people are familiar only with one gas the methane or propane that
they burn for heating and cooking. Some of the non-heating applications of
these gases are discussed in section 3.2.2.
Hydrogen
Molecule: formula HH
weight 2.05
Physical boils 20 K
melts 14 K
viscosity
tonnage cost $150
Reactions: burns with invisible flame
saturation of unsaturated organics
reactions
metal oxides reduced to metal
At first named dephlogisticated air by Priestley, hydrogen was well-known
and fully investigated by the end of the 1700s.
Ethylene
Molecule: formula CH
2
=CH
2
weight 28
Physical boils 169K
melts 104 K
viscosity
tonnage cost $50
Reactions: reactive hydrocarbon
saturation to ethane
HCl, etc., addition reactions with
many chemicals
polymerisation
forms polyethylene ethylene/
propylene at high pressure and
with catalysts
1.4.1 Special gases
The term special gases, as it applies to the industrial gases industry, refers to
any gas prepared and delivered in small quantities, up to a few tonnes or so,
and of an unusual level of purity or in unusual mixtures. The most important
classes of special gases are described here.
Semiconductor gases. These include hydride gases for deposition (like
silane, and chlorine and fluorine-containing gases for etching and
mixtures of these. They all have to be prepared to exceedingly high purity
specifications.
Calibration mixtures. These comprise all those mixtures of gases used for
setting up gas and other analysers. They vary in simplicity from simple pure
gases, in individually checked cylinders, to mixtures of ten hydrocarbons
used for calibrating gas chromatographs in the petrochemical industry. One
of the newest areas of use for these gases is that of calibration for automotive
exhaust gas analysers. Now that most countries have compulsory testing for
engine exhaust, there are tens of thousands of analysers needing calibration
for statutory purposes.
Ultrapure gases. These are needed for many purposes. Any scientific work
needs to begin with ultrapure gases, whilst semiconductor and nuclear work
often needs extraordinary purity.
Rare gases. These are sometimes known as noble gases and are those in
Group VIII of the Periodic Table. Argon is supplied as a regular bulk or
industrial gas but the others are supplied by special gas divisions or
companies in small quantities. Helium and sometimes neon are available in
liquid form, for use as cryogens. In addition to their other uses, all these
gases are used in the lighting industry for filling glass bulbs.
42 INDUSTRIAL GASES
Gaseous chemicals. These are basically those reactive chemicals whose
vapour pressure is such as to require gas-handling techniques. They are
mainly used as intermediates in specialised chemicals such as pharmaceu-
ticals. Gas companies traditionally deliver small quantities of these com-
pounds via their special gases divisions, with large (tens of tonnes plus)
quantities being delivered by the chemical manufacturers direct. Chlorosi-
lanes, phosgene, ammonia, ethylene oxide, propylene, butanes/butenes,
chlorine, hydrogen sulphide, sulphur dioxide, methyl chloride and di-
methylamine are some of the most common examples.
Refrigerants. These are gases which liquefy under modest pressures and
are typically more or less inert and non-flammable (ammonia and ethylene
are examples of gases used as refrigerants which do not fit this description,
however). A range of available boiling points allows refrigerator machinery
to be constructed with a wide range of cold-end temperatures. Although
hitherto chlorofluorocarbons (CFCs) were the most important refrigerants,
because of their bad environmental effects (stratospheric ozone depletion)
they are now fast disappearing. In their stead, hydrofluorocarbons and
HCFCs, such as HCFC-134a, are now growing in use.
1.5 The gases industry
1.5.1 Which gases matter?
It is difficult to find precise figures for the amounts of gases that are used in
industry, so diverse are the uses to which they are put, but the following
figures should give at least the magnitudes involved. The majority of the
gases industry turnover, $10 billion or so (but larger if usage internal to
companies were included), is concentrated in a few dozen key gases. Table
1.5 shows some of the most important industrial gases and their approximate
US production (~30% of world production on average) in millions of tonnes
(various sources). A number of gases are significant in the turnover they
bring because although they are produced in rather small quantities
measured in tonnage, they sell for very high prices, e.g. gases such as argon,
silanes, acetylene, sulphur dioxide and nitrous oxide.
It may seem curious that some of the most important gases in tonnage
terms are not those which might be ordinarily thought of as important in
industry. In fact the list in Table 1.5 is distorted because enormous tonnages
of gases such as CO or hydrogen are used but only a very small percentage of
usage is recorded in statistics. If the full usage were included, then the list in
Table 1.5 would look even more skewed towards gases apparently
unimportant in general industry. However, many of the largest tonnages
are completely consumed within the pipework of a single chemical plant.
For example, ethylene is produced in immense quantities by the cracking
of petroleum. However, in many cases, the ethylene is both produced and
used within the same plant for production, for example of vinyl chloride or
alcohol.
The so-called merchant gases, those produced and then sold to other
companies or industries, are more interesting because of their interaction
with industry at large. Small changes in process economics within the
chemical industry can result in vast changes in the tonnages of non-
merchant gases. For example, the use of ethylene increased about three-
fold in the USA between 1977 and 1981 due to new petrochemical process
changes.
Many of the uses of these internally consumed non-merchant gases are
too specialised to be discussed in this volume. Similarly, heating appli-
cations of fuel gases such as methane and coal-gas, whose use could be
supplanted by other heat sources, are not discussed. The remaining gases
are truly industrial and are the subject of this book. They are used all over
industry and are an irreplaceable component of many and varied pro-
cesses.
Some of these gases seemed esoteric when they were discovered but now
have important practical applications. An example of this is silane
This was made in small amounts from the 1850s onwards but was, as
prepared then, highly unstable, normally spontaneously decomposing or
igniting in seconds after only a few cubic centimetres had been prepared.
Today, silane is no laboratory curiosity but is made, shipped and used in
tonnage quantities as a result of enormous strides in the purity of silane and
carefully designed production and utilisation plant.
44 INDUSTRIAL GASES
1.6 Corporate structure of the industrial gases industry
The major players in the manufacture and supply of industrial gases today
are mostly long established companies with global businesses dating back to
1900 or so. They are large multinational, or arguably global, businesses.
The nineteenth century saw the the peak of colonialism. Technically ad-
vanced nations in Europe and the USA administered or strongly influenced
vast territories in Africa, Asia and the Americas. The gas companies of these
nations expanded using the colonial areas or spheres of influence of their
countries of origin in the early years of the twentieth century, before most
colonial links began to be severed in the middle of the century.
The supply of compressed oxygen gas from air distillation, together with
acetylene, was the root of their business expansion. Compressed oxygen was
needed in modest quantities in the developing economies of all the former
colonial areas. These requirements were a good match to the capabilities of
the growing multinational gas companies. In most cases, the size of the gas
demand was insufficient to be worth the developing country setting up its
own firm, particularly when set against the considerable technical expertise
needed.
The LAir Liquide company of France and the BOC company of the UK
are two of the largest players who fit this picture precisely. Air Products is a
major player established more recently during the 1940s when US wartime
requirements allowed rapid expansion. More regional, but still very large
concerns, are gas companies such Nippon Sanso of Japan and Messer
Griesheim of Germany. Messer Griesheim and the Union Carbide Linde
Gases Division were the relatively small gas divisions of the chemical giants
Hoechst and Union Carbide Company (UCC), respectively. Union Carbide
have recently (1992) split off the gases division as PRAXAIR.
4
Linde of
Germany is now a rather rare example of a basically national industrial gas
company.
In Japan, gas for steel, until recently the biggest gas application, has often
been manufactured by the steel companies themselves. This has meant that
Japanese gas companies have been smaller than multinationals from the
West. However, companies such as Nippon Sanso have grown much bigger
in recent years and now have many ventures offshore. Nippon Sanso and the
other Japanese companies still have some way to go, however, before they
reach the global scale of operations achieved by the majors.
4
The Union Carbide name is linked in the public mind with the Bhopal chemical plant accident
of 1984 (see Jones, 1988). Particularly for a company supplying gases, the term gases often
having emotive poisonous overtones, this is bad. In 1990 a proposed UCC plant in Scotland
attracted so much bad publicity that the company withdrew the proposal despite the fact that
central Scotland already has two companies operating gases plants and has massive
petrochemical works of much greater safety import. The name change to PRAXAIR may well
benefit UCC.
Many of the gas majors have very significant non-gases operations. BOC
has substantial interests in vacuum pumps, distribution of refrigerated food
and other products, and in hospital healthcare areas such as anaesthetics and
medical equipment. Air Products and Chemicals, as its name suggests, is
also a large speciality chemical supplier, e.g. in fluorinated chemicals. AGA
has a large engineering wing, whilst Linde of Germany is a major
manufacturer in Germany and the UK of fork-lift trucks.
There appear to be many more gas companies than there really are
worldwide because of the habit of many of them of retaining the old name of
companies long ago merged or taken over. For example, in the USA,
AIRCO and Liquid Air turn out to be none other than BOC and LAir
Liquide respectively. Many comparatively small regional gas suppliers have
been taken over by the handful of multinationals in recent years, again
further reducing the number of alternative suppliers: in most markets, in
effect, industrial gases are an oligopoly.
In several small newly industrialised countries, there are joint venture
companies formed between the gas majors. For example, the Malaysia
Oxygen Company is a joint venture between BOC and LAir Liquide. In
other countries, such as Turkey, there is a joint venture between a large
conglomerate local company involved in other industrial activities with a
multinational gas supplier: in effect, a marriage between local knowledge
and influence, and multinational technology and know-how.
1.6.1 A typical industrial gas company
Although the industrial gases business is largely an oligopoly, the small
number of companies involved do conform at least partially to a common
pattern. A typical industrial gas company:
is multinational;
has a significant non-gas operation (see above);
is not R&D (research and development) intensive, with 2% or less of

turnover in R&D for its gases business: thus industrial gases lie half way
between general engineering (~1%) and research-intensive sectors, like
semiconductors and pharmaceuticals, which have >10% of turnover in
R&D;
has a significant distribution business: except for pipeline business with

very low gas prices, all gases sold must be distributed to the user; the costly
specialised cryogenic tank trucks and cylinder filling plants all add to the
size and importance of this aspect of their operations;
has a fairly high capital intensity: large and expensive production units,
the specialised transport fleet and the need, because customers rarely
supply their own, for a large inventory of cylinders (and cryogenic tanks)
all dictate rather large capital requirements;
46 INDUSTRIAL GASES
has an average to high turnover per employee: although the turnover per
employee is large in gas production, distribution of gas, especially the
cylinder business, is labour intensive;
has an average growth rate: although industrial gases is a mature industry

in most respects, there are still significant growth areas in the application
of gases; as a supplier to the shrinking heavy industries, gases sees some
losses but in relatively new areas such as water treatment and food
preserving there is strong growth; similarly, there are compensating gains
and losses geographically, with fast-growing South-East Asia economies
providing growth to compensate for losses elsewhere.
A typical industrial gas company is internally organised at least partly as a
matrix. The matrix has tonnage, bulk, compressed and special gas business
units along the x-axis. These business units have largely separate staff for
marketing, sales and finance, but also use the services of semi-autonomous
central service departments. Departmental functions such as engineering,
gas applications development, R&D, analytical laboratories, purchasing,
stores, personnel and so forth, form the units along the the y-axis (see Table
1.6).
There is furthermore, within the operation of many of the companies, an
additional degree of matrix operation in that a particular project may draw
on staff from various functions and from different business units.
The picture is further complicated by the multinational operation of the
gas industry. World centres for financial services, legal services and R&D
are operated by most of the gas companies. The world central R&D facility
is where expensive research and analytical tools like electron microscopes or
mass spectrometers are usually located. Basic topics in gas science
technology as it applies to production and safety, and long-term work, are
carried out centrally. The central facility is complemented by small local
R&D laboratories. The small local laboratories concentrate on the appli-
cations of gas and the specific requirements of local users of gases.
1.6.2 Monopoly in the gas industry
There are frequently, and occasionally with some justification, cries of
effectively only one supplier, whilst in others there may be two or three who
appear to collude on prices. Even within the massive economic area of the
European Community monopoly investigations have been carried out on
gas supply.
The basic problem is that gas is difficult to transport long distances at low
cost and gas plants are subject to scaling laws giving big savings for big
plants. In addition, in areas of intense industrial activity there are pipeline
networks for common gases like nitrogen. The supply of industrial gases,
except for special gases, is thus technically most efficiently operated as a
monopoly. However, a monopoly is most undesirable from the point of view
of fair and free trade, and governments have sought to try to reconcile the
conflicting technical and commercial requirements. Perhaps the best
compromise is an oligopoly combined with the ready availability of
government controls if necessary.
In the UK especially, recent years have seen the denationalisation or
monopoly e.g. the network-based telephone, electricity, water and (meth-
ane) gas companies. In each case, along with the privatisation went a new
but small government body whose job is to regulate the new near-monopoly
private companies to ensure fair-trade practices, i.e. to inject the influences
which competition would have supplied.
Whether these regulator bodies and the privatisation programme is
successful will be a judgement for history. Their effectiveness is currently the
subject of controversy with many of the new privatised utilities at least
temporarily making large amounts of profit for their employees and
shareholders, relative to much higher risk manufacturing companies.
However, perhaps these could provide a model for regulation of the
industrial gas industry, allowing it to exploit fully the economies of scale but
not to abuse a near-monopoly position.
1.6.3 Gases industry supply chain
The supply chain of many industries can be represented as a supply chain,
rather like a food chain in ecology. Just as in nature, the food chain
becomes a pyramid when the value of the business at each level is
monopoly against the gas companies. In some smaller countries there is
privatisation of many national utility organisations who have a natural
48 INDUSTRIAL GASES
considered: whilst supplying industrial gas plants might be a billion dollar
industry, supplying the gases themselves is worth ten billion or more and the
applications that use them are themselves worth hundreds of billions.
At the top of the gases industry food chain are the suppliers of gas
production plant. Some of these are also operators of the plants that they
make; they are vertically integrated. Air Products, Linde, and BOC are
examples. Other gas plant makers are independent in the sense that they do
not normally operate their own plant and do not offer merchant gas for sale
but seek to profit mostly by selling plant, e.g. Costain Petrocarbon and
Hitachi.
In parallel with these plant suppliers are the makers of components such
as cryogenic storage tanks, compressed gas cylinders, valves and fittings.
Again, some gas suppliers also make components, for example AGA and
BOC make tanks. In the main, however, components come from specialist
companies and specialist divisions of larger firms. Gardner Cryogenics is a
large US specialist in cryogenic tanks, for example, and TI Chesterfield of
the UK makes huge numbers of hot forged steel cylinders, a business that
requires particularly massive and expensive specialised equipment.
Many valves and fittings used in the gases industry are little different from
those in general use for flow control in methane, fuel gases and compressed
air. However, high pressures in gas cylinders and low temperatures in
cryogenic tanks have led to special design components for critical places.
Cylinder valves and pressure regulators are completely specialised to
compressed gases, for example, as are the long-stemmed valves used for
liquid nitrogen and oxygen. Specialised companies, and special divisions of
larger general flow control component companies, have grown up to meet
these needs.
Below the plant and component makers in the food chain are the gases
suppliers themselves and below the gases suppliers are the users of gas. Just
as some gases suppliers are vertically integrated with their suppliers, some
gases suppliers are downwardly vertically integrated with users of gases.
Linde, for example, supplies industrial refrigeration plant which is a large
user of freon and other industrial gases, while in the 1980s, the King Harry
pizza business, a large user of liquid nitrogen for freezing, was part of BOC.
1.6.4 Geography of gas supply
Industrial gases cover the complete range of commodity types studied by
geographers. Some are highly specialist, valuable and shipped to customers
around the world from a single producer in small compressed gas cylinders.
At the other end of the scale are gases produced by large on-site plant,
produced by taking services such as power and cooling water from the
customers plant, and delivered solely to that customer a few yards away
through a pipe.
Tonnage gases are low-cost gases in large volumes such as the oxygen in
steelworks. Tonnage gas is frequently taken from one point of production to
one point of consumption by a large diameter pipe, although in areas of high
industrial development there are networks of piped gas, especially nitrogen.
Silicon Valley, south of San Francisco, and the industrial areas of northern
France and Belgium have extensive pipes for tonnage industrial gases.
Plants are attached on the pipeline in various places and provide continuity
of supply by acting as each others back-up in the event of plant failure.
Customers are attached to the pipeline where necessary. By using inex-
pensive steel (black iron) pipes of moderate pressure capability, and with
the aid of regional assistance funding in some cases, pipelines have been
extended over considerable distances at an economical price. With moder-
ate and fairly constant flow rates in many industrial processes, there has
been no need for the dramatic cycling of pressure up to 200 bars or more,
seen in methane fuel gas supply pipelines. Similarly, the effective length of
the pipe is such that intermediate pumping is not needed.
and transferred to a bulk tank. Some gases (e.g. CO and can be
delivered in compressed form in trailer loads: an articulated semi-trailer
fully loaded with 20 or so tonnes of manifolded cylinders. Bulk gas is
transported only regionally; above 100 miles or so the transport costs begin
to rise and favour supply from a closer plant. Similarly, the costs of common
compressed gases, which are delivered by truck up to a 100 miles or so, are
composed mainly of transport costs at larger distances and of handling costs
at short distances.
It is notable that there is considerable variation in gas costs geographi-
cally. The cost of carbon dioxide, for example, is several times lower in the
USA compared to Europe. This large discrepancy probably reflects large
investment in the extraction of from waste streams in the USA. In
Europe there are fewer supply sources.
5
In some parts of Europe, is in
effect a local monopoly because a single large source within the chemical
industry is the only economical local source. (Waste streams in the chemical
industry often provide a copious and inexpensive source of There are
national and regional differences in electrical power costs and these impact
tonnage prices greatly, bulk prices to some extent and compressed gas prices
only to a very minor extent.
For gases delivered in cylinders, the costs of labour and transportation
completely dominate the business and these vary widely from country to
country. For bulk gases, prices vary steeply with distance from the nearest
plant, with troughs in price near concentrations of plants and peaks in
between. In general, the location of plants reflects the location of users.
5
In the UK, for example, with few low-cost sources, carbon dioxide is imported by a special
cryogenic tanker ship to a large holding tank in London.
Bulk gas means gas delivered by truck, usually in cryogenic liquid form,
50 INDUSTRIAL GASES
However, historical reasons (for example, the movement of the steel
industry into fewer sites) mean that gas plants are not ideally located. In the
case of gases like sources of raw gas are rarely ideally placed for
consumers. Sources are largely in the older steel and chemical industries,
whilst many users are in the newer disciplines of food processing and
semiconductors.
Smaller gas companies. In amongst the large multinational gas companies,
although these companies dominate the business, there are a number of
smaller operations. These fall into a number of categories:
(i) The Mom and Pop operations seen widely in the USA, where a small
family business delivers cylinder gases, the business often being about
primarily propane and butane for heating. The firms also act as agents for
the large gas companies for other gases for welding and other purposes, and
try to build this up as it is a less seasonal business than heating fuel gas. This
sort of firm is a significant feature of most developed countries gases
market. However, the USA market relies on these small local concerns to
the extent that most of the large industrial gas players there reach the
majority of their cylinder gas customers via these firms,
(ii) Small special gases businesses typically manufacture only one or two
gases themselves, relying on bulk supplies from the larger gas firms and large
chemical companies for the other 50 or 100 gases they might typically offer.
The bulk of their activities are cylinder filling, mixtures, analysis and
purification of ultrapure gases, together with an inevitably large amount of
effort on sales and shipping. The smallest of these businesses are typically
national in extent and concentrate on only one sector, e.g. say calibration
mixtures, medical uses or electronics uses. The larger of these businesses are
international and the very largest offer ranges wider than the large
integrated industrial gas players.
(iii) Local bulk/tonnage gas firms typically might have a supply scheme
taking pipeline gas, with some bulk liquid delivery by road tanker to nearby
customers. These firms are becoming increasingly rare as the large gas
players take them over. Some of them had historically grown to be large
companies, e.g. the Big Three Corporation which was a substantial firm,
centred on bulk gas in a small area in southern USA, until its takeover by
AIRCO at the end of the 1980s. A more unusual example of this kind of
company is the UK firm Hydrogen Supplies which delivers in tube-
trailers in north-west England and surrounding areas.
(iv) Chemical companies by-products, for example the Distillers
business of ICI in the UK until its takeover by Hoechsts gases subsidiary
Messer Griesheim. Distillers took by-product from ICIs chemical
plants and distributed it in liquid form in road tankers. The fashion for
focusing on core activities' which swept the UK and US corporate world in
the 1980s led to many of these businesses being sold off either to large gas
firms or to the in-place set of managers running the business. (Some of these
latter management buyouts seem to have proved very successful. How-
ever, some seem to have been a convenient way for a large firm to make
large numbers of employees redundant at reduced cost or with reduced
impact on their own operations, as the new buyout firms typically end up
with a much smaller staff.) To a large extent, this hiving-off of non-core
businesses is still happening today (1996) and by-product sales and delivery
activities are still regarded as non-core in many industrial sectors.
1.7 Financial features of the gases industry
The capital intensiveness of gas production plants has already been
remarked on. With a typical on-site cryogenic air-separation unit (ASU)
costing several million dollars, and installations at steelworks ASUs often
amounting to $50m and upwards, this might appear to need little justifi-
cation. It should be remembered, however, that many other industries are
now rivalling the chemical and steel industries in their concentration of
capital: semiconductor wafer fabrication plants, for example, are now
being built with single-site investments of over $1bn.
Projects to produce or apply gas are subjected to financial scrutiny which is
similar in general terms to that used in all of industry. Breakeven,
discounted cash flow and internal rate of return are all regularly calculated.
There are differences, however, between gas projects and the general run of
industry. Because the gas production processes tend to be capital intensive,
some figures are perhaps debated more intensely in the gas industry than in
other industries. There is usually (section 1.7.3) a big incentive to build a few
very large plants rather than a larger number of smaller ones, although
distribution costs also have to be considered. This is not typically seen in the
chemical industry where, with small numbers of very large customers, the
distribution costs will typically be low. Capital expenditure on gas produc-
tion is large and must typically be accompanied by generous levels of
guaranteed or almost guaranteed sales, a factor which favours the oligopoly
seen in the gases industry.
There is a big difference between the return expected in the application of
gases and that expected in the production. An application for using gas
might have a payback of less than 2 years and an effective IRR of 30% per
annum. (The IRR, internal rate of return, is the effective rate of interest that
a company pays on interest it has invested in itself.) However, most gas
production projects have long paybacks (5 years or more), high breakeven
(often over 80% of capacity) and low IRR just a few percent above
inflation. One result of the low level of maximum returns, the low upside,
has been that means have arisen to minimise the downside (the risk) of
52 INDUSTRIAL GASES
production projects. The take-or-pay clause is a typical example (section
1.7.2).
1.7.1 Product exchange in special gases business
Another result of the capital intensiveness of gas production plant is that
special gases are often only manufactured or extracted in a relatively small
number of locations.
At one end of the scale, some special gases are simply mixed from gases
readily available in bulk from many sources. Many calibration gases based
on air gases and hydrocarbons are examples of this. At the other end of the
scale are gases used in the semiconductor industry, such as silane or arsine.
These are only made in bulk at half a dozen locations worldwide in high
purity grades. The more exotic compounds, such as organometallics, are
quite frequently only available from one supplier. Even a relatively common
gas such as helium is only available from a few dozen locations where its high
concentration in a large natural gas field makes extraction economic.
European He comes largely from two US plants, for example, sup-
plemented by the Polish plant operated by BOC.
The necessary scarcity of sources, balanced with the huge multiplicity of
uses for these gases, means that special gas suppliers must rely on a huge and
complex network of product exchange. A gas may quite typically be sold
three or four times before reaching its end-user.
1.7.2 Take-or-pay and argon credits
In large tonnage gas contracts there are a number of unique features,
relative to standard industrial practice. The take-or-pay contract attempts to
match the gas companys commitment to a large and expensive plant to the
revenue from the customer. Such a contract stipulates that the customer will
pay for a mi ni mum quantity of gas, no matter whether that gas is actually
consumed or not. There is a good industrial logic behind take-or-pay since it
enables the cobbler to stick to his last. An expert gas supplier can operate a
plant for a customer who would not be an expert in operating that plant and
wi l l have more time to pursue his own business. Without take-or-pay, the
risk for the gas supplier, with a heavy capital expenditure, would be too
great and the user would have to run his own plant.
Take-or-pay commitments from the gas user often have corresponding
commitments from the gas supplier to supply-or-pay. Gas suppliers then
build-in storage tanks and vaporisers so that they have back-up supplies of
liquid gas which can be vaporised during periods when the plant is
undergoing repairs and maintenance.
Argon credits are unique to the gas industry. In the past, industrial gas
suppliers operating large air-separation plants used to extract a small
fraction of the argon from the oxygen stream, with modest profits from the
argon sales resulting. Now, however, there is a much larger demand for
argon and argon is efficiently extracted and sold to a large and highly
profitable market.
In a typical tonnage contract the gas supplier is operating an air-
separation plant supplying gas over-the-fence to one customer. That
customer will now typically insist on an argon credit. This means that the
customer pays a price for his oxygen or nitrogen reduced by an amount
reflecting the quantity of liquid argon that the industrial gas company has
been able to extract. The potential value of the argon credit is seen if it is
remembered that bulk liquid argon can sell for 500 per tonne or more,
whilst oxygen prices run as low as 20 per tonne or less. Even with argon at
5% of the oxygen stream, the argon from an air-separation plant could be
worth more than the oxygen. However, most industrial gas companies
would baulk at allowing the argon credit to exceed the oxygen price and
paying their customers to take the gas.
In addition to specifics such as argon credits, there are other more generic
features in gas industry contracts which are unusual, although not unheard
of, in other areas of business. These are often associated with the necessarily
long-term nature of the gas companys involvement with its customer. For
example, it is common to sign-up contracts of 5, 10 or even 15 years. Since
the value of money cannot remain stable over this kind of time, it is
necessary to include in such contracts formulae for changing the gas price
charged. These formulae are based on indices of industrial prices for power,
raw materials, capital equipment and labour. Such long-term agreements
have to be carefully written to be equitable to both sides, even in the event of
unforeseen future economic upsets, and have to avoid falling foul of the
1.7.3 Scaling-up gas plants
A gas production plant is most efficient at a particular product output and
will tend to be less efficient when operating at below that output. It will also
be less efficient at above that output, if more output is possible. As a result of
this, plants of many different sizes are required to match the many different
demand sizes. It is therefore important to be able to estimate the cost of
building a plant of any size.
To do this a cost-scaling law is applied as follows
where is the scaling power.
In this case, the scaling power applies to the whole plant. However, it may
be more accurate to calculate the new cost taking into account different
scaling powers applying to different components. In other words, taking
anti-trust or anti-competitive laws which exist in most countries.
54 INDUSTRIAL GASES
as the new cost and as the old, scaling powers
etc., for subsystems x, y, etc., and the capacity ratio new/old to be R,
i.e.
For example, the cost of controls has a scaling power of 0. Thus the cost of
the new capacity plant controls will be the same as the old capacity plant.
(The same controls will be required, more or less irrespective of the size of
the plant.)
For a reactor bed, the scaling power is 1, and the cost of the new capacity
plant will be increased or decreased exactly in proportion to its size increase
or decrease. If twice as much bed material is required, then the cost will be
roughly twice as much. (For very small plants, this rule may break down, as
bed material suppliers may have higher prices for small quantities.)
The scaling power for a pressure vessel is also 1, because the thickness of
the walls of the vessel must increase with the size of the vessel (section
2.4.7). The scaling power for a non-pressure vessel is only two-thirds, since
only the area of the tank wall needed will increase as the square of the linear
tank dimension, whilst the volume will increase as the cube (Table 1.7).
The approximate scaling rules for the components shown in Table 1.7 are
easy to estimate and can be used as a guide in the absence of good price
information and to interpolate where price information is scant. Scaling
rules for many other components where a scaling rule is not at all obvious
can be set up by empirically reviewing prices published and paid for similar
types of equipment. Some examples of where this has been done are given in
Table 1.8.
Tables of component costs of the sort shown in Table 1.8 are available
from some trade associations, and are published in some chemical engineer-
ing references (e.g. Perry, 1984).
The scaling laws are distorted from what might be expected because of the
existence of subsidiary scaling laws:
in purchasing, bulk buying reduces costs;

there is always a tendency to gold-plate equipment which normally has a

zero scaling law, such as controls or instrumentation: equipment that is
desirable but might be considered disproportionately expensive for small
plants will tend to be included on a bigger plant;
the timing of purchases will affect costs: some components, such as bed
materials, for example, can be bought after the rest of the plant is
completed in a big plant, whilst others must be bought near the beginning
of fabrication;
market distortions often affect the very largest pieces of equipment of a

particular type because there will often be fewer suppliers of them.
Finally, there are various discontinuities in the cost of plant:
when the plant exceeds the size that can be transported by road, more
on-site work will be needed and this is more expensive than factory work;
reliability and back-up considerations;
when plant exceeds handling equipment maxima: if simple, inexpensive

and easily hired cranes can be used this minimises costs.
A combination of these scaling laws, weighted according to their
importance in a particular project type, lead to the scaling law for that type
of plant. Scaling laws can also be derived directly if enough similar plants
exist and prices are known, and this has also been done. For example, Perry
(1984) gives an exponent of 0.65 for steel plants with outputs in the range
1504000 ktonnes per annum. Common examples of chemical plant
generally fall in the exponent range 0.6 to 0.7. This arises from combining
the scale factors of, for example, heat exchangers or furnaces of 0.8 to 0.9
with scale factors for gas compression and pretreatment of 0.55.
Impact of costs on design choices. Decisions about whether to use gases in
compressed, bulk liquid or tonnage are not always straightforward. A
tonnage installation, for example, requires a large capital investment in
plant or possibly pipeline, whereas an equivalent bulk liquid installation will
require only modest capital investment but will have a much higher running
cost. A large firm which can borrow money easily at low cost and is able to
56 INDUSTRIAL GASES
make long-term commitments will prefer the first approach, whereas a small
firm with a rapidly changing market will prefer the second option, even if it
means considerably higher overall costs.
Similarly, demand patterns have a considerable impact on supply routes.
High peak demand, with low average consumption will point towards bulk
storage rather than a production plant, while continuous or near-continuous
users will be biased towards production plant. The possibility of operating
production plant on low-cost night-time electrical power may also bias users
towards production plant rather than bulk gas delivery.
However, as a guideline, the smallest economical cryogenic tanks contain
a few tonnes of liquid, and would normally be used for gas usage of around a
few tens of tonnes per annum. Below this amount (a few tens of thousands of
cubic metres) it is wasteful to use cryogenic tanks. There are many small
liquid installations, however, down to 2001 liquid cylinders. This is because
the handling of many gas cylinders is the maximum capacity) is
tedious and because the cost of gas in compressed cylinders is of the order of
five times more than bulk liquid gas.
The scaling laws for cryogenic air-separation plants (and many other gas
production plants) mean that the production of less than a few tens of tonnes
per day is uneconomic. Over the last decade this breakpoint has been
blurred to some extent by the advent of pressure swing absorption (PSA)
plant. For users needing only moderate purity nitrogen or oxygen with a
steady demand, PSA plant offers a high capital/low running cost alternative
to bulk liquid gases from the smallest to the largest bulk liquid user.
The financial value of experience. It is a platitude that experience is worth
its weight in gold but few attempt to put an actual monetary value on
experience. However, many long-term surveys of industrial production
have resulted in the experience curve (Figure 1.12). The experience curve
is generally reckoned to be a logarithmic one. The unit cost of producing a
gadget is reduced by a multiplying factor A, for example each time
the total number produced is, for example, doubled. Mathematically, this is
where is the initial cost. Sometimes the curve is plotted in terms of profit
per uni t , in which case it is a curve tending asymptotically to a maximum if
price is constant.
The experience curve can actually be used to put a value on experience in,
for example, assessing the value of a company's technology during a merger.
If a company has to start again from square one, then it will have to climb up
the experience curve.
Similar curves have been drawn for gas production and similar effects are
seen. For many gas production plants the technology is now well-established
and the asymptotic efficiency and unit cost is being approached. The
Figure 1. 12 The experience curve.
well-developed nature of plant technology is seen in the high up-time
figures claimed by the manufacturers. Up-time, the proportion of a gas
plants lifetime that is spent either producing gas or ready to produce gas, is
now claimed to be better than 99%, whilst many larger plants with
redundant systems and back-up claim 99.9% or better.
1.8 The key role of gases in industry
Industrial gases pervade industry today. In a country of 50 million or so
people, like the UK, there will be well over a hundred thousand different
locations where gases are delivered. Every day well over a million tonnes of
industrial gases, non-fuel gases, are consumed in industry worldwide. The
ubiquitous role of gases in industry can be illustrated by looking at a complex
consumer product and tracing its components back to their origins to see
how gases were used in their manufacture. For example, the gases involved
in the manufacture of a pocket calculator are listed in Table 1.9.
Use of gases in the manufacture of a spraycan or aerosol. The role that
industrial gases have taken in the production of many modern devices is now
considerable. The aerosol or spraycan is a packaging technology, for
example, which embodies gases in almost every single component:
Propellants. Propane and isobutane mixtures are the mainstay of
aerosol propellants in the 1990s. CFC compounds went out of fashion for all
but medical applications and some paints as ozone layer concerns grew.
Nitrogen, and other compressed permanent gases have been used to
propel liquids out of aerosol cans since the 1960s and are increasingly
popular.
58 INDUSTRIAL GASES
Formulations. Many aerosol formulations include gases actually in the
dispensed mixture or in the manufacture of the ingredients. Materials
included in mixtures vary from carbon dioxide or nitrous oxide to
hydrocarbons or dimethyl ether. For example, shave gels include a small
percentage of isopentane, a compound which exists as a liquid misicible with
the shaving gel whilst under pressure inside the aerosol can. However, when
the gel is released from pressure and warmed up as it is dispensed onto the
hand, the isopentane boils and bubbles, converting the concentrated gel into
a lubricious foam ideal for shaving.
Can fabrication. The tinplate aerosol can will have been produced by an
oxygen-based steel process, whilst the steel may well have been rolled and
heat-treated in a hydrogen or other bright-annealing atmosphere prior to
electro-tinning. The aluminium aerosol will have been made from alu-
minium electrolytically extracted at high temperature with an argon
atmosphere.
The plastic and rubber parts of the aerosols valve and seals may also have
been processed with gas-based processes: moulding machines cooled with
liquid nitrogen or tumbled in a liquid nitrogen deflashing operation. The
plastics themselves may derive from ethylene, propylene or butadiene
gases, polymerised on their own or with other compounds.
1.8.1 Smokestack and leapfrog economies
A vital parameter in the design of new cryogenic air separation units is the
ratio of nitrogen to oxygen produced. Market studies carried out before such
plants are constructed have therefore been very interested not just in the
quantities of gases which a given market might need but also in the ratio of
oxygen to nitrogen needed. Observers in the gases industry have noted that
this ratio actually varies between different countries, depending on the
nature of their industrial base.
A newly industrialised country will be carrying out relatively large
amounts of steel fabrication (cutting and welding) and will tend not to have
the luxury industries of food processing and semiconductors. These
emergent economies tend to use a lot of oxygen and much less nitrogen. An
example of one of these emergent industrial economies would be a smaller
African nation or a poorer Asian nation such as Bangladesh.
Smokestack economies, based predominantly on large-scale heavy
industry, have an oxygen to nitrogen ratio around 1:1. A smokestack
economy will have enormous steel industry applications, requiring nitrogen
and argon as well as oxygen (fabrication of oxygen will be only a small
proportion of requirements), while other metal processes such as heat
treatment will be using nitrogen. In addition, the smokestack economy will
have a bulk chemical industry, using nitrogen in fair quantities, so as well as
requiring much greater quantities than the emergent economies, the
smokestack economy ratio will be closer to 1:1. In the USA in the
early was roughly 1:1, while today countries such as India have
large steelworks and fabrication industries but a large and growing sector in
services and high technology, so they also have a ratio of 1:1
Today the mature industrial economies of the USA and Europe have a
ratio of 1:2, reflecting both the decreasing amount of heavy industry,
such as steel, using oxygen and increased use of nitrogen in the service, food
and high-technology sectors.
A recent phenomenon are the leapfrog economies. Some newly
industrialised countries on the Pacific Rim are jumping straight from an
emergent stage into high-technology manufacturing, such as the semicon-
ductor industry, which use large amounts of nitrogen. They have a low
ratio of 1:4, similar to the ratio in air. The gas consumption ratios for
different economies are given in Table 1.10.
emergent economies tend to use a lot of oxygen and much less nitrogen. An
Smokestack economies, based predominantly on large-scale heavy
60 INDUSTRIAL GASES
Detailed studies of the local demand pattern for gases are necessarily
added to this overall market view before air-separation and other plants are
built. Most often, a merchant gas operation will be run as a piggy-back
operation on a major pipeline gas supply plant, such as an oxygen plant
supplying a steelworks. The scaling laws of cryogenic plant particularly
favour this approach.
2 Gas technology
2.1 Separating air gases
2.1.1 Introduction
It may seem surprising that any energy at all is needed to separate mixtures
of gases; after all, the molecules of gases are not attracted towards each
other so the separation process does not have to exert any force on the
molecules. However, the act of mixing gases together increases disorder, or
entropy, and energy is always required to reverse entropy. The theoretical
minimum energy required to physically separate an equal mixture of two
similar gas molecules is about This is much less, by a factor of
a hundred or more, than the energy of typical chemical reactions. This, then,
is the fundamental reason why physical separation processes for gases, even
rather imperfect physical separations, are favoured over chemical gener-
ation.
Thermodynamics of gas separation. Consider a perfect separator, where
an input mixture of equal parts of A, B, C and so forth is pressurised into a
chamber lined with membranes (Figure 2.1). One membrane perfectly
passes A, one B, one C and so on. To produce products at one atmosphere
absolute pressure, the feed must be supplied at an absolute pressure equal to
the number of products.
For a mixture of two components, such as air, the energy required is thus
that needed to compress air under ideal conditions (section 1.3.1) to 2 bara.
Hence
energy for separating mixture =
The energy needed for non-equal parts of gases is reduced, so that air
separation into nitrogen and oxygen requires only However, if
only the 20% oxygen of the separated gas is required, then the actual
separation costs of a 50/50 mixture are doubled to whilst
oxygen cost will rise to (Table 2.1). The relevant equation for
the separation energy per mole of product from a mixture of a% A and
b% B is
62 INDUSTRIAL GASES
Figure 2.1 Theoretical model of air separation.
It is perhaps surprising that the extraction of even a rare gas such as xenon
does not theoretically require a particularly large amount of energy.
6
Unfortunately, the ideal semipermeable membranes required to carry out
the trick shown in Figure 2.1 do not exist, although imperfect membrane
6
It is interesting to note that in principle energy is available free simply by extracting methane
from the air: methane combustion generates around far more than is required to
extract it. Also, is surprisingly easily extracted from the air, a fact not used in any chemical
processes unless atmospheric used as a reactant in the setting process of lime-based
mortars counts as a chemical use.
GAS TECHNOLOGY 63
separation is possible, as we shall see. Also, the ideal isothermal com-
pressor, needing only the above energy, can only be approached in large
sizes.
This, then, is the simple thermodynamic science of gas separation. The
achievement of separation is more complex than this simple science might
suggest, however. Since the beginnings of efficient gas separation tech-
nology by cryogenics in the late 1890s, many different configurations and
cycles have been devised. Before going into some of those cycles, however,
it will be useful to look at some of the components which will inevitably be
part of them. For example, whatever method is chosen to carry out gas
separation, a gas compressor is generally used chosen to energise the
separation.
2.1.2 Gas plant building blocks
Introduction. Generally speaking, specialist firms are used to fabricate the
items which are common between the gas industry and other industries. It is
still essential, however, to know the basic characteristics of these building
blocks in order to be able to design a complete plant most effectively. The
manufacture of gases involves the use of equipment which often differs
significantly from hardware used in many other chemical and industrial
processes. The hardware used can have a bigger impact on the final plant
cost and efficiency than the configuration of that hardware in the plant.
A glance at any diagram for gas separation (Figure 2.13) will reveal that
vital parts of all cryogenic plants are compressors, heat exchangers and
perhaps turbine expanders, as well as the core of the plant, the distillation
column, and a host of smaller items such as valves and a nest of high-quality
pipework. Gas compressors for cryogenic work are some of the largest units
seen and may vary in output pressure from a few bar to a hundred bar or
more to satisfy the thermodynamic requirements of the cycle used.
Because the gas from the compressor will be cooled so strongly, the
presence of condensibles such as water and oil which would freeze and clog
the plant must be reduced to infinitesimal levels. Whilst water is often
contained in the feed gas or air and must be specifically separated, most
compressors for cryogenic work must also not allow any oil into the outlet
stream. This is in contrast to many room temperature applications, where
the presence of a trace of oil mist is actually desirable for lubrication of
machinery downstream. The consequences of allowing oil into a cryogenic
plant are so dire that practically all compressors for this use are oil-free
designs. Non-cryogenic plants also have to have efficient compressors.
Pressure swing absorption (PSA) plants are occasionally built in very small
sizes where convenience and low purchase cost rather than efficiency is the
prime requirement but the economics of most PSA plants still depend on
efficient compression.
64 INDUSTRIAL GASES
Heat exchangers and insulation for cryogenic service are similarly
specialised. In a plant operating above room temperature the efficiency of a
heat exchanger is not quite so important as in a cryogenic plant. A heat loss
or leak of 1 kW into the cold parts of a cryogenic plant will require 3 kW of
electrical power (assuming the highest possible Carnot efficiency for the
refrigeration in the plant) to reject it at 77 K, or 70 kW at liquid helium
temperature. By contrast, a 1 kW loss or leak from a high-temperature
process will require only 1 k W compensation with direct heating (less with a
heat pump). Hence cryogenic heat exchangers and thermal insulation have
to be more efficient than is normal.
In addition to this emphasis on efficiency, the range of materials out of
which cryogenic heat exchangers can be fabricated is limited. Simple and
inexpensive carbon steel fabrications are ruled out because steel embrittles,
whilst copper is expensive and difficult to weld. In practice aluminium alloys
are the mainstay, being strong, weldable and with good thermal conduc-
tivity. Stainless steels such as 304 or 316 are also widely used where higher
strength is needed and despite their low thermal conductivity they, like mild
steel in high-temperature work, can still be made into effective heat
exchangers for some services.
Designs also differ for cryogenic use with the need to provide mi ni mum
temperature difference between incoming and outgoing streams while still
providing low pressure drop. Plate-fin types dominate, providing high
efficiency, with the possibility of balancing heat flows from several high- and
low-pressure streams of different mass flow rates being easily satisfied in the
versatile plate-fin design. Again these differ from normal practice and the
large effect on plant performance means that heat exchangers for gas
separation deserve special discussion. (Insulation is discussed in section
2.4.)
2.1.3 Compressors
The trapezoid symbol for compressor in a process diagram hides the details
and subtleties on which tens of thousands of engineers have worked for a
century or more to perfect. Efficient, reliable gas compression with low
plant capital cost still requires wise selection on the part of the process
designer and skilful engineering on the part of the compressor designer.
A compressor is characterised by its throughput, compression ratio and
efficiency. One might add, since compressors are often the least reliable
single unit in many gas systems, that the mean time between failures
(MTBF) and % up-time should also be considered as vital characteristics.
The throughput is usually given as gas volume at normal temperature and
pressure (NTP), whilst the compression ratio is simply the ratio of the outlet
over inlet absolute pressures. The efficiency measures are less standardised
but comparison with an ideal adiabatic compressor is useful.
GAS TECHNOLOGY 65
Some of the fundamental difficulties of gas compressors in the region of
often used for air separation can be summarised as
follows:
axial compressors cannot achieve high enough pressures except in the

largest sizes and with huge numbers of stages (at 10 MW or above axial
compressors become competitive)
centrifugal compressors can only achieve high pressures in large sizes and
then only with multiple stages

reciprocating compressors have many wearing parts and suffer from poor
reliability, efficiency and up-time except in small sizes
7
gases at high pressures easily leak out of loose seals but tight seals involve
wear problems and frictional losses

many compressors cannot use any seals and use very small clearances and
high speed. Slight distortions of rotor or stator in this case can lead to
failure

oxygen at high pressure will encourage ignition of many seal and
constructional materials, particularly if friction causes local heating

all air compressors must efficiently separate out water from their outlets:
compressing air causes water to condense out (section 2.1.5)

screw and reciprocating oil-lubricated compressors must efficiently
separate oil, water and air in their outlets.
An interesting variation on the positive displacement types of compressor
is the water-ring compressor. This looks at first glance like a centrifugal
compressor, being a central set of vanes rotating in a housing. However, the
pumping action is provided by the vanes being mounted eccentrically in the
housing, whilst a rotating ring (actually a hollow cylinder) of water
provides the seal and piston in the design. The water ring is held in place by
the high centripetal force. Water-ring pumps can be designed with large
clearances between the mechanical parts, yet still provide good compression
ratios. Designs such as the water ring are useful for feed gases loaded with
particulates, which would cause excessive wear in standard compressors.
Water-ring and similar compressors, however, must separate the water or
other working fluid from the output even more efficiently than types which
have only a lubricating fluid present, which adds another potential
complication.
Figure 2.2. shows two compressor types: centrifugal and screw. The large
centrifugal compressor from Sulzer is a five-stage compressor with inter-
cooling using tube bundles (three of these can be seen in an upright position
in the figure).
7
Reciprocating compressors can, however, last a very long time: one factory the author worked
in is still using a large reciprocating compressor made by a Paris firm 90 years earlier for
commercial cylinder filling of special gases.
66 INDUSTRIAL GASES
Figure 2.2 Compressor types: (a) centrifugal (courtesy of Sulzer (UK) Ltd); (b) screw.
Axial and centrifugal compressors. Axial compressors are typically suit-
able for large-scale compression to moderate pressures, i.e. a few bar. A
single stage is capable of only a few hundred millibar or so pressure rise,
beyond which blade tip sealing and aerodynamic stalling occur, so that any
practical axial compressor requires several or even dozens of stages. The
very largest plants may use axial compressors up to high-pressure but
require massive installations with many dozens of stages.
Centrifugal compressors. The centrifugal compressor is suitable for
GAS TECHNOLOGY 67
medium- to large-scale compression to a few tens of bars. It is efficient and is
common at larger air-separation units. The pressure rise achievable is
limited by the tip speed and the air density. The maximum pressure rise
in a centrifugal compressor with tip speed working on air can be
estimated as approximately
where k is a constant for air of about This follows simply
from considering an element of gas at equilibrium inside the fan while it is
whirling around, i.e. with the fan in the stalled condition.
The pressure achieved is a very steep function of the fan tip speed. To give
some idea, a tip speed approaching the speed of sound is needed,
or so, to give compression ratios of better than 2:1 or so and 3 bara. For large
compression ratios, centrifugal compressors are normally cascaded, with
coaxially mounted units all driven from one shaft and aerodynamically
designed smooth housings or volutes to carry the gas efficiently between
stages.
Because of the steep function of tip speed, and limits on the rotational
speeds of machinery, very high speed and large size must be achieved for
adequate compression, limiting these compressors to relatively high volume
requirements, i.e. above
Rotary compressor seals. One of the problems of all compressors at high
compression ratios is that the compressed outlet gas leaks past the shaft seal
between the motor and compressor stages. Compressors with a compara-
tively small compression ratio can in effect simply accept leakage by using a
labyrinth seal. This is no more than a set of grooves on shaft and housing
which forces the leaking gas to pass though a tortuous path of annuli.
For high compression ratios at least, however, sealing is a problem and
much ingenuity is often devoted to its solution. The standard system of the
stuffing box seal, where lubricated fibrous ring seals are pressed around the
shaft by an adjustable collar, is rarely acceptable. The stuffing seal is more
efficient if the collar forces it down hard onto the shaft but then power is lost
through friction. Rubber seals are often used, using carbon filler to reduce
the coefficient of friction which would otherwise be high in the absence of
oil, and elastomer composition chosen for compatibility with gas and
lubricant. These are often tensioned with internal metal springs or use a
shape which allows operational gas pressure to aid their task. Carbon seals
are also possible and have advantages in being self-lubricating. A small
leakage is often designed into seals, in effect allowing the seals some gas
lubrication.
Yet other compressors avoid problems by using what might loosely be
termed active seals. In these designs, a face seal with a pneumatically
regulated thrust ring is employed. The seal is closed by a spring pushing the
68 INDUSTRIAL GASES
Figure 2.3 Schematic of a pneumatic thrust ring seal with a mechanical self-balancing action.
two faces together. In operation, after start-up, the high pressure gas from
the compressed side compresses the spring and opens the seal slightly,
allowing gas to drop in pressure down the narrow annulus around the shaft
and closing the seal again. The seal takes up an intermediate position with
the leak rate (and shaft to seal clearance) regulated by the spring strength
and the size of the annulus supplying gas from the high-pressure side.
Some designs of sprung plastic or metal rotating annular seals also achieve
a feedback effect, as the two-part annular spring seal is wedged open by the
high-pressure side but as soon as the leak through the seal increases,
lowering the pressure inside, the spring tightens the seal. Figure 2.3 shows
how such a pneumatic ring seal with a mechanical self-balancing action
works. By choosing the spring pressure correctly the leak can be regulated to
an acceptable level, whilst achieving the very low friction of a non-contact
seal. Both these active seals have the feature, often an advantage, that they
provide an almost hermetic seal on the shaft when it is stationary or at slow
speeds. However, low-speed operation can lead to excessive wear. Similar
arrangements are also seen with oil seals but these are rarely seen in air
separation unit (ASU) compressors because of the potential for oil
contamination.
There is an interesting variation often used on centrifugal compressors
which avoids many of these problems with seals: the motor can be operated
pressurised with the outlet gas. Yet another way of avoiding the seal
problem is to use magnetic coupling. In this the pump is driven by a magnet
inside a pressurised diaphragm, while a corresponding magnet outside the
diaphragm is driven by the external motor. Clearly, when the drive on the
outside is an a.c. electromagnet system, then the magnetic coupling system
amounts to the system above.
GAS TECHNOLOGY 69
Figure 2.4 Centrifugal compressor flow versus discharge pressure showing surge line.
Compressor surging. Occasionally, typically during start-up or an unfore-
seen accident, centrifugal and axial compressors can suffer large surges up
and down in revolutions, accompanied by large variations in output pressure
and vibration. Surging is a result of the interaction between a compressor
and the process load it is put under.
Looking at Figure 2.4, it can be seen that if flow is decreased a little from
point A, then pressure builds up a little. This is stable, since the increased
pressure will tend to increase the flow rate and bring the system back
towards to A again. If, however, the flow is decreased past point B as a result
of throttling in the downstream process, then a slight decrease in flow will
lead to a decrease in pressure, which leads to a further decrease in flow,
which leads to a further decrease in flow, and so on. The final effect is that
the flow from the compressor will collapse towards C, when the load on the
compressor motor falls, so that it can accelerate until the rotational speed
has built up sufficiently, perhaps to 3200 rpm to raise the pressure beyond
the pressure at B at zero flow. Then flow begins again, slowing down the
compressor back to its 3000 rpm norm. This cycle manifests itself in a
hunting or surging effect on the compressor speed and is highly undesirable
because vibration levels are raised and fatigue damage to critically stressed
rotating parts and welds in the static parts is increased.
Surging can be dealt with, particularly in larger systems, by bypassing
some of the compressor output if the surge point B is approached. This
dump valve assembly increases the flow through the compressor and
reduces the pressure achieved, forcing the compressor to go back down its
characteristic curve towards A on Figure 2.4. The dump valve can be to a
vent line or back to the compressor inlet.
A pressure vessel downstream of the compressor is another way of dealing
with the problem of surging. A comparatively small pressure vessel installed
70 INDUSTRIAL GASES
downstream of the compressor will suppress surging and also provide a
buffer of compressed gas to handle sudden downstream demand changes
whilst allowing the compressor time to adjust.
Reciprocating, oil-injected screw and gear pumps. Reciprocating com-
pressors are suitable for extremely high pressures or low volumes at
moderate pressures. The simple piston and cylinder linked to a rotating
crankshaft with non-return valves on the inlet and outlet is still the
commonest design. A single-cylinder compressor, however, has a very
unsteady output and needs to be linked to a large pressure vessel for
applications requiring continuous output. Multi-cylinder arrangements with
suitable valvegear enable fairly steady pressure and flow to be maintained.
There are many wearing parts in a reciprocating compressor: piston seals
and skirt, cylinder walls, bearings of piston and crankshaft, and non-return
valves.
Simplified designs linking a piston directly to a reciprocating drive
mechanism, such as an electric solenoid, are available in small sizes.
Because they eliminate the crankshaft and several rotating bearings they
offer some advantages. There are even small non-contact reciprocating
compressors available which mount the piston and drive unit on springs. The
springs are arranged to allow a large axial movement but to control radial
movement accurately. For low compression ratios, where the necessary
clearances are acceptable, the piston can then move in and out of the
cylinder without touching the cylinder walls. Such units are used in
spacecraft and similar applications where ultimate reliability is needed.
Reciprocating compressors are relatively simple to make and maintain,
and are generally cheap to purchase. However, they have drawbacks in the
form of high maintenance and reliability problems and are generally avoided
in industrial installations today.
Screw compressors are now a commonly used alternative to a reciprocat-
ing uni t . Screw compressors can achieve a few bar (or a few tens of bars with
multiple rotors) in a simple unit which can be made in sizes down to a few
tens of cubic metres per hour. They give a steadier output than a
reciprocating unit, with fewer moving parts. At their heart are two
multi-thread screws with very exact profiles which interlock: a male profile
on one and a female profile on the other. The screws or rotors are often
sychronised, likes Rootes and Northey blowers, by gears mounted on their
shafts.
Screw compressors rely on the carefully machined screws fitting together
precisely to fine tolerances, however, and are potentially sensitive to
damage from clashing of the rotors in the event of dust contamination or
temperature excursion. They are of very high efficiency with only very small
losses from leakage from output to input, at least at lower compression
ratios. To achieve higher compression ratios (10:1 for air, typically) they are
GAS TECHNOLOGY 71
normally operated with continuous injection and recovery of oil. The oil
enables the rotors to act as their own synchronizing gears and reduces
leakage losses as well as making a more forgiving, lubricated assembly. The
injected oil is recovered in a cyclone and a coalescing filter and then cooled
prior to reinjection, the oil thus acting as an additional coolant. Special oil
must be used which does not easily emulsify as this would lead to
accumulation of water from the feed in the recirculated oil. Oil-injected
screw compressors can give longer life, better efficiency and improved
reliability relative to the reciprocating devices they generally replace.
Older in origin and usage than the screw compressors is the gear pump.
This relies on one or more pairs of gears which enmesh gas on the inlet side
and carry it to the outlet. The simplest possible gear pump is the Rootes
blower, comprising two gears with two teeth on each gear, which is thus
dumbell or lobe-shaped in outline. The Rootes blower and Northey claw
forms of gear pump rely on external synchronising gears. All gear pumps
need accurate tooth form and very small clearances to function correctly on
gases and are easily contaminated with particulates. However, they can
achieve very high compression ratios in small sizes. They are used where
very small flow rates are needed and on gases such as propane or butane
which are handled as liquids.
Cooling compressors. Compressors do not invariably need cooling. If the
subsequent process required high temperature, as well as high-pressure gas,
then cooling may not be appropriate. However, compressors, or at least
single stages of them, are approximately adiabatic devices. Gas compressed
will rise in temperature in accordance with the simple adiabatic rule
thus raising air pressure from 1 bar at temperature 20C to 10 bar will cause a
temperature rise to 293C or so. This is not what is required for
low-temperature cryogenic apparatus and is also quite unsuitable for
membrane or PSA separation, which require near room temperature
operation. Moreover, the energy expended in raising the temperature can
rarely be usefully employed and even if it were it amounts to electrical
heating of gas, which is rarely economic. Isothermal compression is the
desirable ideal.
Adiabatic temperature rise. The rise in temperature in an adiabatic
compression may be visualised as follows. When a piston is forced down on
an assembly of gas in a thermally insulated container, the motion of the
piston adds to the momentum of gas molecules that hit the piston during the
process. This leads to a general increase in the speed of the molecules, which
is seen as an increase in temperature.
72 INDUSTRIAL GASES
Figure 2.5 Multistage intercooled compressors approach to isothermal performance.
The simple well-known result for an ideal adiabatic gas compression or
expansion is
where g is the ratio of the specific heats of the gas at constant pressure and
constant volume, i.e. This in turn leads, when the ideal gas law is
added, to the given formula for adiabatic temperature rise.
The optimum gas compressor is an isothermal one. This, as discussed in
section 1.3.1, requires an energy of to arrive at pressure from input
absolute pressure
An adiabatic compressor, were it a single-stage device followed by a
cooler, would require more energy
for an ideal diatomic gas.
The energy has gone into heating the gas up, in this case to as much as 293C.
The optimum approach for compressor efficiency is via multistage
compressors with interstage cooling (Figure 2.5). This is the closest
approach to isothermal performance possible for practical equipment.
Inlet gas cooling is another option for improving compressor efficiency. If
a source of cooling is available, from a waste stream perhaps, then further
improvement can be offered, perhaps even exceeding isothermal perform-
ance from ambient. Finally, it should be remarked that some larger
compressors require careful control of their cooling: sudden changes in feed
GAS TECHNOLOGY 73
air temperature, outlet flow rate or cooling water can cause the compressor
to change in temperature too rapidly. When this happens, some components
cool or heat more quickly than others and there is the potential for problems
of clashing or even seizing occurring, given the close fits used.
JouleThompson cooling. Expansion across an orifice or valve in a pipe is
the standard versatile source of refrigeration in a cryogenic plant, utilising
the JouleThompson (JT) effect. A JT expansion causes cooling in gases
below a certain temperature (the inversion temperature) in the following
way.
An ideal gas does not cool on expanding across a nozzle: its pressure is
reduced but it does no work on any engine and, from considerations of
conservation of energy, remains precisely at the same temperature.
However, most real gases have a small attractive force between their
molecules and they are able to do work against this cohesive force and lose
energy, thus cooling, on expanding across a nozzle. The JT coefficient
describes this effect by giving the temperature (in C) by which the
temperature drops on throttling a gas through 1 bar of pressure drop. The
JT coefficient varies with pressure and temperature becoming, in general,
negative at higher temperatures (above the inversion temperature) and
smaller or negative at higher pressures (Table 2.2). The JT effect is thus not
a powerful one: for nitrogen at room temperature compression to 20 bar will
yield only a 4C lowering in temperature.
For nitrogen, the temperature at which the JT coefficient becomes zero is
348C. For lighter gases, this inversion temperature is lower, and these gases
must be precooled before JT expansion is useful: hydrogen is at 200K,
whilst helium is 43 K. For gases where data is not to hand, extrapolation of
JT effects can be achieved by noting that the JT coefficient at
constant enthalpy, is given by the fundamental thermodynamic formula
where is the thermal expansivity of the gas, i.e. at constant
pressure. As is simply 1/T for an ideal gas, as expected, an ideal gas shows
74 INDUSTRIAL GASES
However, if the van der Waals coefficients a and b are known for a
gas, then an approximate can be calculated which at least shows the
correct qualitative behaviour and can be used for extrapolation. For
example, for a gas at moderate pressures with van der Waals b coefficient
small, the JT coefficient is given approximately by
As expected, is zero if the van der Waals a coefficient is zero and increases
at lower temperature.
Expansion turbines for adiabatic cooling. The low efficiency of the
JouleThompson cooling effect and the fact that it is unavailable for many
gases near room temperature means that there is a requirement in many
cryogenic plants for another source of refrigeration. High-pressure cold gas
is often further cooled by an adiabatic expansion. The cooling effect
achieved is easily predicted as the converse of the adiabatic heating
calculated above for gas heating in compressors, although some corrections
will be necessary due to the low temperature of operation leading to
non-ideality of the gas. The calculation is more complex where the gas
actually condenses partly to liquid on expanding.
Whatever the difficulties of calculation, achieving adiabatic expansion at
low temperatures is not straightforward. Some early air-separation plants
employed, and some modern helium plants still employ, reciprocating
expansion engines, which operated in a very predictable way. However, to
achieve necessary throughput of gas such engines rapidly become very large
and unwieldy if they are not to have to run at impossible rotation speeds.
Add to this the difficulties with sealing the piston to cylinder at such low
temperatures, lubricating the whole assembly and operating valves at high
speed.
The efficiency of adiabatic expansion turbines in cooling is calculated as
follows. If the efficiency of the engine is there is a perfectly efficient
compressor and cooling heat transfer from to is required,
then Carnot efficiency for the whole cycle, i.e. gas compression + expansion
operating as a refrigerator, is given by (= work theoretically needed/
work actually needed) as follows
This follows from considering a perfect Carnot cycle heat engine but adding
the effects of an expander efficiency of by increasing the heat flow to the
hot reservoir by Examination of the formula above shows that if the
turboexpander has no work output at all, all its work is dissipated as heat in
the gas stream itself, then the efficiency of the process sinks to zero and there
is no cooling effect at all, it is simply a throttle valve. However, this assumes
an ideal gas and for most gases there is a JouleThompson cooling effect.
GAS TECHNOLOGY 75
The efficiency of an expansion turbine grows as the output temperature is
reduced until, roughly below 100 K or so, it is as efficient to use a turbine as it
is to use an evaporating liquid cycle refrigerator. The latter, the standard
refrigerator, is most efficient when required to deliver cooling at the
temperature of its evaporator and at temperatures close to room tempera-
ture. The turbine has no defining optimum temperature of cooling efficiency
in the same way and becomes steadily more efficient as temperatures are
reduced. Isalski (1989) gives the figures shown in Table 2.3 for power
consumption of three different routes to 40C, using a complete cycle of
compressor, heat exchanger and expansion route.
Today, an expansion turbine is the standard means of achieving adiabatic
cooling. It necessarily has to run at exceptionally high speed, as it must offer
a high pressure drop to give a significant cooling effect. If the turbine is
overbraked it will become inefficient with too much gas simply leaking
through its blades or around its blade tips at the high differential pressure
needed and, as noted above, function simply as a JouleThompson valve.
The speed requirement makes turbines technically difficult although it has
the benefit of making the machine surprisingly small even for large
throughputs. Typically single-stage or double-stage centrifugal turbines are
used. These run at speeds from 20 000 to100 000 rpm, the high speeds being
used in smaller units. The exotic turbines essential for efficient liquid helium
production have to run at much higher speeds, up to nearly 1 million rpm,
although the turbines that run at these heroic speeds are very small, only a
few millimetres in diameter.
Expansion turbines often employ gas bearings, at least in smaller sizes, to
achieve high-speed efficiently. Oiled bearings have the problem, along with
ball or roller bearings, that unless insulation is assured (large machines
having an advantage here), the low temperatures can cause the oil to freeze
or the rollers or balls to fail via fatigue, causing the bearing to seize.
Standard journal bearings suffer from whirl at high speed, especially
with gas lubrication. This is where the shaft, as well as revolving, orbits
inside the bearing. Whirl is suppressed by providing a damping force by
76 INDUSTRIAL GASES
arranging gas inlet orifices circumferentially, along with exit orifices which
have a small volume above them, around the bearing journal. The small
volume gives a damping action thus suppressing whi rl . Thrust bearings can
also employ an externally fed gas-bearing principle but more often use
self-acting spiral groove bearings. These provide a pumping action which
creates their own gas film but have the snag of not giving any lubrication
when stationary or at low speed. Magnetic bearings have also found
application in expansion turbines.
A magnetic bearing can be a passive bearing, that is, a powerful
permanent magnet in the bearing repels a powerful permanent magnet on
the shaft so that the latter can rotate freely. However, passive bearings are
limited. First, they cannot provide a complete suspension system; for
example, another type of bearing must stabilise the shaft in the axial
direction if the passive bearing provides stability in radial directions.
Second, there is a limit to the force that can be practically obtained with the
constraint that at least one of the permanent magnets must be on the shaft.
Active magnetic bearings rely on an electronic feedback signal. A shaft,
which need not be permanently magnetic, is suspended by attraction to a set
of powerful electromagnets, amplifiers varying the current through the
electromagnets in order to keep the shaft central. The movement of the shaft
in axial or radial directions is detected typically by an electromagnet-based
proximity device, providing the input to the suspension magnet amplifiers.
Active magnetic bearings can provide full stabilisation so that a shaft need
have no physical contact with its bearings or housing.
Expansion turbines sometimes have to deal with a significant amount of
wet gas two-phase flow, i.e. partially condensed liquid/gas, and must
thermally isolate the cold turbine wheel as effectively as possible from the
shaft carrying away the power (Figure 2.6). The power generated by the
turbine is often simply dissipated in an oil brake but in large units power is
used to generated electrical power or to operate a compressor co-mounted
on the shaft.
RanqueHilsch vortex tube. The vortex tube achieves, like a turbine, a
cooling effect additional to that of the Joule-Thompson expansion but
without a conventional expansion engine. In a vortex tube air expands at a
nozzle tangentially arranged to one side of the divider in a tube with a central
divider (Figure 2.7). Cold air is extracted on the opposite side of the divider
from a hole in the middle of it, whilst hot air is collected from the nozzle side
of the divider from a annular exit at the end of the tube.
A vortex tube achieves cooling via expansion at the nozzle at the entry to
the tube but achieves an overall effect on part of the gas stream better than a
JT valve by separating a cool gas stream in the middle of the tube from the
rest of the gas, which exits the tube having slowed down and heated up. In
effect, work is done by the cold part of the gas stream to heat up the hot part
Figure 2.6 Single-stage expansion turbine for small-scale liquid nitrogen use (courtesy of
Linde).
78 INDUSTRIAL GASES
Figure 2.7 RanqueHilsch vortex tube.
of the gas stream. The precise theory is incomplete, however, as, because
practical vortex tubes operate with a noisy, unsteady flow, this has not yet
been modelled. Hilschs original work used an orifice of 2.2 mm in a 4.6 mm
tube supplied with up to 10 barg pressure air and showed temperatures as
low as 35C, from room temperature. Later work improved the achievable
temperature drop to 65C and increased the fraction available as cold air up
to 30%. Even with these latter figures, however, efficiency is only 10%
Carnot.
2.1.4 Thermal transfer in gases: heat exchangers
Heat exchangers are ubiquitous in gas manufacture and application. In
many industrial situations the performance of the heat exchangers is vital to
the overall efficiency of a plant. In many applications of gases heat
exchangers are necessary to raise or lower the gas to the appropriate
temperature of a process or perhaps to stabilise a process against ambient
and gas temperature variations. At the most fundamental level, many
medium-size gas applications use cryogenic liquefied gas because it is the
appropriate form of gas for medium-scale gas supply but actually require
room temperature gas. They use simple heat exchangers called vaporisers to
vaporise liquid gas and warm it to near room temperature.
A typical situation in industry is where a process is to take place at an
elevated temperature but the product gas is needed back at room
temperature. If the feed gas at room temperature is warmed up using simple
fuel or electric heaters and the product is cooled with tap water, then the
process will be very inefficient. The energy used in the warming process is
expensive, as is the copious quantity of cooling water needed. Except for the
GAS TECHNOLOGY 79
very smallest plant, it will be much more efficient to employ a heat
exchanger or heat exchangers, ideally in counter-current mode. In this way
hot product exiting the process transfers heat to the incoming cold feed gas,
warming it up to near the process temperature. In liquid air distillation
plants (section 2.1.5) the process is run at very low temperatures and the
emphasis on efficiency is much greater but the principles are the same.
With an ideal counter-current heat exchanger, one that would have a zero
temperature difference across it or differential pressure in either
process stream, a process could be operated at any temperature with no loss
of energy. In practice, in the region of 1 to 10 K is tolerated as a
compromise between capital and running cost. A small value is in
general only achieved by a larger and more expensive heat exchanger, whilst
a larger value wastes energy but can be achieved by a smaller, cheaper
exchanger.
Forced convection under turbulent conditions is the most common
situation where gases are to be heated or cooled strongly and efficiently. The
equations for heat transfer under these circumstances are usually very
complex. However, the use of dimensionless groups simplifies equations
and allows the setting of operational zones in which reasonably accurate
empirical formulae can be used. Whole books are devoted to heat exchanger
design for gases, and specialist firms make the units to order. It is interesting
nevertheless to consider here some of the possible designs and design
considerations.
Exchangers must clearly separate two fluids: the gas being cooled or
heated and the working fluid, whether it is liquid cryogen, Freon refrigerant
or another gas. The fluids may be at different pressures, as well as different
temperatures, and may in generally be potentially corrosive. Nevertheless,
the walls of the heat exchanger must be thin so as to maintain the best
possible thermal contact.
The choice of high thermal conductivity materials, such as copper or
aluminium, is standard although, perhaps surprisingly, inexpensive carbon
steel or cast iron can occasionally be used at higher temperature. The
strength and chemical resistance of 316L-type stainless steels and the ease of
fabrication of small-bore pipes using 316L stainless steel means that these
are also popular, despite having thermal conductivity five or ten times worse
than aluminium. The decrease in thermal conductivity seen in many metals
and alloys works against the designer for heat exchangers needed at liquid
air temperatures, with many aluminium and stainless steel alloys almost a
factor of two lower in conductivity at 77 K relative to room temperature. It is
important to note, however, that different alloys, different batches of alloys
and pieces of the same batch differently mechanically worked or heat-
treated can show thermal conductivity which will vary by a factor of two or
more, with the greatest variations typically below 100 K.
The geometric design of heat exchangers varies enormously. Designs are
80 INDUSTRIAL GASES
compromises between providing more intimate contact between fluids via
smaller interleaved channels and providing a lower pressure drop via large
channels. The straightforward tubes-in-shell design originally seen in steam
engines is still popular. For very large units, or those with an arduous service
(large rapid thermal cycling, high pressure), a development of the
shell-tube design is occasionally seen in cryogenics in the form of the helical
wound-tube design.
More common in cryogenic practice, however, are welded or vacuum-
brazed assemblies of cross-flow or counter-flow plate-fin heat exchangers
(Figure 2.8). These are usually more compact and inexpensive for a given
service. Plate-fin exchangers, once complete, are large rigid assemblies but
have the disadvantage of being more susceptible to unrepairable cracks due
to thermal cycling. Differential thermal expansion considerations also limit
the maximum possible size of plate-fin units. Aluminium contracts about
0.4% (copper and stainless steel 0.3% for comparison) between room
temperature and liquid nitrogen temperature, which is more than the yield
strain so distortion of components which are unevenly cooled is likely even
where they do not crack. Pipework attachments to heat exchangers and
other components running in cryogen must always be designed to allow free
thermal movements to avoid problems.
It turns out that the design of a whole heat exchanger can be summed up in
the simplest case by the same equation used above (Searls bar equation) for
each small elemental area, dA, of that exchanger by using an effective
average temperature difference. This effective average temperature differ-
ence in a heat exchanger can be shown by integration, with assumption of
constant U value, to be given by the log mean temperature difference
The choice of counter-flow in heat exchangers yields a smaller exchanger
unit for a given service, as the effective temperature difference is smaller in
that arrangement (Figure 2.9). In the idealised example chosen, a certain
heat exchanger is compared in performance in counter-current and co-
current mode. The achieved final temperature of the hot stream is 223 K in
co-current mode and 211 K in counter-current. This reflects the fact that in
the co-current case the temperatures in the exchanger vary exponentially so
that the small heat transfer at the outlet, where is small, is not
compensated by the large heat transfer at the inlet, where is large.
In the counter-current case the temperature distribution is still exponen-
tial but the distribution is almost flat, giving a higher average heat
transfer. Counter-current mode of operation is the only possible mode
where the hot stream must be cooled below the cool stream outlet
temperature or where the cool stream must be heated above the hot stream
outlet temperature, as shown in Figure 2.9. It is surprising how well large
counter-current heat exchangers can work, particularly in larger sizes where
GAS TECHNOLOGY 81
Figure 2.8 Plate-fin heat exchanger (courtesy of IMI Marston Ltd).
heat leaks to the outside are less significant. The rather small amount of
cooling from the JouleThompson effect can easily overcome the inefficien-
cies in the heat exchanger system in air liquefaction. If this were not the case,
air separation by liquid-air distillation would not have been possible in the
early days, using only JouleThompson expansion, and even modern plants,
with their turbo-expanders giving powerful cooling, would not be so
successful commercially.
Passages in a heat exchanger will in general be related to the flow required
at reasonably low differential pressures. Liquids will achieve the same mass
flow in much smaller channels, for example, so liquid channels will be
smaller than gas channels. The need to allow two-phase flow, boiling and
condensation further complicates design.
82 INDUSTRIAL GASES
Figure 2.9 Stream temperatures in (a) a co-flow exchanger and (b) a counter-flow exchanger.
A typical calculation for a heat exchanger is
where Q is the heat transferred, dA is a small area of exchanger and
is the temperature difference. The heat transfer coefficient U for a
particular design of heat exchanger is usually quoted for different liquids and
gases (with different pressures quoted for gases), U can be estimated if this is
not the case by adding together the reciprocals of thermal conductance of
the exchanger wall and the two film heat-transfer thermal conductances
Under typical forced convection conditions approximate numbers for these
GAS TECHNOLOGY 83
can be looked up. For example, for atmospheric air, values of around
100 W are typical.
The equation
accounts for the two film conductances and the wall, of thickness t and thermal
conductivity k. The film conductance is so low that in some applications it is
unnecessary to use a particularly good thermal conductor for fabricating the
exchanger. When the thickness t is not required to be unreasonably thin, and
provided the material satisfies thermal expansion considerations, simple steel
exchangers can be used. Aluminium is typical, however, and allows for very
compact heat exchangers which are physically very strong.
Suppose an air stream of 1500 at 300 K and 10 bara is to be
counter-current cooled against a 1000 m
3
h
1
nitrogen stream at 100K and
10 bara. Let us assume that a heat exchanger design is used where U is, say,
100 W overall. What size (area) of exchanger is needed and what
outlet air-stream temperature might be expected? If the simplified
assumptions are made that specific heats of the gases are constant and the
same, and that U remains constant over the temperature range, then (see
Figure 2.9) with counter-current design and an air stream achieving 211 K,
the log mean T is 85 K.
area needed
Clearly, in order to achieve acceptably low T values, large areas of heat
exchanger are needed. Hence the need for the convolutions of the plate-fin
designs.
2.1.5 Distillation of air
Distillation of cryogenically cooled liquid air at a few tens of bar or less
pressure is still, nearly a century after its large-scale use, the technique of
choice for preparation of large quantities of oxygen, nitrogen or argon.
The atmosphere: raw material for a $20 bn industry. The typical composition
of the earths atmosphere at sea level, expressed as fractions of dry air by
volume, is as shown in Table 2.4. The amount of water vapour in the
atmosphere is frequently forgotten in calculations, but it can be substantial on
a warm, humid day.
8
8
The saturated vapour pressure of water varies with temperature approximately exponentially
(section 1.3.14), being 1 bar at 100C and 0.023 bar at 20C. The latter implies that on a day with
high, near 100%, humi di t y, the atmosphere wi l l be about 2.3% water. On compressing to 10 bar
the air will still only have the capability of carrying 0.023 bar of moisture. However, 0.023 bar is
now 0.23% of the air stream so the other 2.07% will condense in the outlet of the compressor. For
every 100 compressed, in this case 1.8 1 of water will be condensed out. This means that a
compressor of quite modest capacity, say a few hundred cubic metres per hour, will require a
hosepipe to lead away the condensate.
84 INDUSTRIAL GASES
Pre-purification and liquefying of air. Before air can be liquefied it must be
free of impurities which would block up the plant. As well as filtering out
particulate matter, any gases or vapours which could freeze up and cause a
blockage at 100 K will need removing. Also any inflammable impurities
need to be removed because most inflammable gases will accumulate in the
liquid oxygen and may, when mixed with it, form an explosive mixture.
(Several early liquid oxygen plant explosions were caused by the latter
problem.
9
)
Pre-purifiers may employ PS A or temperature swing absorber systems, or
they may use the RHE (reversing heat exchanger) system, which freezes out
impurities. Temperature swing absorbers (TSAs) are the norm. Beds of
silica gel, zeolite or carbon (section 2.1.6) are selected for their abilities to
absorb moisture, carbon dioxide, methane and acetylene at room tempera-
ture. This they do for a considerable period before being regenerated by a
flow of hot gas. Two beds are employed so that one bed can be regenerated
by the heated dry nitrogen stream whilst the other is on-line.
RHEs are often noisy, a fact which is of importance unless the plant is in
an isolated location. They function by presenting a cold plate to the cool
compressed air passed through them. Moisture and carbon dioxide, the
principal impurities, are condensed. After a few minutes, the heat
exchanger is purged with a short flow of dry, warm, waste gas, which warms
up and drives off the absorbate in seconds, enabling the plate to be used once
again. The purge gas, released at large flow rate, is usually noisy, even when
it is high up at the top of a vent stack.
Liquefaction of air is still the key technology of modern industrial gases. It
is surprisingly simple, once good insulation and the efficient heat exchangers
described above can be taken for granted. All that is required is for
9
Acetylene, unlike methane and other light hydrocarbons often seen in air, is substantially
insoluble in liquid oxygen. It is this which leads to the problem because if acetylene is
concentrated above its solubility, which is only a few parts per million, it can precipitate,
forming a slush with the liquid oxygen whi ch is in effect an explosive. The precipitated acetylene
formed in t hi s way is exceptionally dangerous.
GAS TECHNOLOGY 85
Figure 2.10 Simplified scheme for liquid air production.
compressed air to be cooled against vapours from the liquid air in the process
and then expanded. The expansion is most efficiently accomplished by an
expansion turbine which allows the cool, compressed air to expand
adiabatically to become a mixture of liquid air and cool vapour at lower
pressure.
Although efficient in cooling and a potential source of some power, the
expansion turbine has to run at around 80 K at very high speed in a
two-phase fluid. These are relatively arduous conditions and these turbines
are expensive and potentially unreliable. Smaller and simpler plants expand
the compressed air through a nozzle or valve using the JouleThompson
effect. Despite the small temperature drop it gives, with efficient heat
exchange the JT effect is quite sufficient for air liquefaction. A scheme of
apparatus using JT expansion, of the type used around the turn of the
century, is shown in Figure 2.10.
Distillation of liquid air. The distillation of any mixture of similar
compounds relies on very simple principles. Basically, when the mixture is
boiled, the vapour will contain more of the element with the higher vapour
pressure. The scale of liquid-air distillation is very large, however. Plants
have been constructed which distil up to 10 000 tonnes of air and more per
day. A plant of this capacity will have a distillation column 6m or so in
diameter and 35 m tall. The separation of nitrogen and oxygen from air is
more difficult than, for example, whisky distillation, however, because the
molecules have a close boiling point, a greater purity of product is expected,
particularly for nitrogen, and because of the cryogenic temperatures.
Multiple distillation is therefore required and is conceptually performed
in a single fractionation tower or distillation column. The column simply
86 INDUSTRIAL GASES
Figure 2.11 Composition diagram for distillation.
integrates a set of retorts and condensers. Liquid air is added to the column
centre. The lower boiling nitrogen is concentrated in the stage above, whilst
the higher boiling oxygen is concentrated in the stage below. The condenser
and reboiler are arranged to provide a reflux flow of liquid and vapour, in
addition to that required for product output, in order to ensure that
equilibrium is approached in all of the stages. The performance of a
distillation column, and the number of stages needed, can be estimated by
graphical methods on composition diagrams such as Figure 2.11.
Composition diagram theory. The composition diagram arises in the
following way. If Raoults law is assumed to be followed, then the vapour
pressures and liquid mole fractions will be related as follows
at boiling point
where is ambient pressure, and are the saturated vapour pressures
of compounds a and b, and the and are the liquid mole fractions
If Daltons law of partial pressures can be used, then
and similarly for Substitution of an exponential type dependence for
and and extraction of X
a
from the first equation leads to and as a
function of temperature.
It is used by tracking across Figure 2.11 from the bottom line to the top
line once for each stage of distillation. A liquid mixture starting at X on stage
1 will distil a vapour of composition X. If this is condensed on stage 2 to
liquid and evaporated, a new purer product Y evolves. IfY is condensed on
stage 3 to liquid and evaporated, it leads to vapour of composition Z, and so
on.
GAS TECHNOLOGY 87
In this way the purity which a certain number of stages will achieve can be
judged. Numerical design methods which automate this process are now
used.
Note that the simple theory above, what might be dubbed a perfect
mixture is only relatively common. Equally common are composition
diagrams which show a peak or trough in pressure due to intermolecular
forces. On distillation these have a composition(s) which cannot be
separated by distillation at a given pressure; the vapour has the same
composition as the liquid being boiled. In these cases, products may be only
partially separated, pure A from azeotrope or pure B from azeotrope, or the
distillation pressure may be changed to avoid azeotropic conditions.
In the simplest arrangement for liquid-air distillation, the column has a
liquid-air feed in the middle of the column from which the heavier
oxygen-rich vapour is stripped of its lighter fraction in the descending or
stripping section and the lighter nitrogen-rich vapour migrates upwards,
becoming purer at each stage, in the upper rectification column.
The progress of each product along the composition diagram lines shown
is followed only when little product is withdrawn. In other words, the reflux
ratio (liquid mass flow/vapour mass flow) is equal to one. As soon as
appreciable product is withdrawn or, equivalently, as less vapour is
condensed for reflux and/or less liquid reboiled, the system will follow a
somewhat different course.
First consider the rectification section. Here the reflux ratio is less than
one and the lighter fraction of the liquid condensed for the next stage should
be reduced, reflecting the fact that vapour containing mostly lighter fraction
is being withdrawn so that the fraction of the lighter component in the liquid
condensed is effectively reduced.
Similarly, in the stripper section, the reflux ratio is greater than one and
the heavy fraction of the liquid condensing on the next tray down is not so
great as before, reflecting the withdrawal of liquid rich in heavy fraction
from the reboiler end of the column.
It can be shown that, under simplifying assumptions, the effect of product
feed and withdrawal points can be modelled by a graphical method involving
the plot of an equilibrium curve of the fraction of A (the most volatile
constituent) in the vapour as a function of the fraction of A in the liquid .This
is the McCabeThiele approximation and it is well-described in Perry
(1984). Similar but more complex constructions can be used for multiple
feed and product streams, although numerical simulation is now used for all
practical designs.
There is a minimum reflux in the upper column (or maximum reflux in the
lower column) beyond which an infinite number of stages would be required
to effect separation. Practical plants with finite columns must operate with
reflux in excess of this amount, typically by 10100%. Clearly, operating
costs increase with increasing reflux, more work is need to the heat pump
88 INDUSTRIAL GASES
Figure 2.12 Single-column nitrogen separation unit.
system, but capital costs increase with increasing numbers of trays in the
distillation column. The best compromise will depend on the relative costs of
power and fabrications.
Single column oxygen and nitrogen distillation. The idealised distillation
column has a feed of mixture to the middle and two pure products but
requires an external heat pump, the refrigerator coil in the condenser and
the heater coil in the boiler, to recycle heat from the condenser to the boiler.
This heat pump requirement can be very large in the above arrangement;
even where a reasonable amount of product can be withdrawn it is inefficient
in practice and very expensive on capital installation.
It is very much more efficient to employ the large main air compressors as
the energy source to drive the process and provide refrigeration. This can be
done by feeding liquid air to the bottom of the column, with some of the
liquid from the bottom being taken off to be flashed off to cool the
condenser. This results in a pure nitrogen stream at the top and a
nitrogen-poor stream at the bottom, i.e. an rectification system or
nitrogen generator system (Figure 2.12).
One important advantage of single-column N
2
units, in contrast to the
otherwise more efficient Linde double-column unit, is that the nitrogen
stream produced is at a pressure a little less than that of the feed air
compressor. A converse trick can be used to make pure oxygen at the
bottom and oxygen-depleted air at the top, i.e. an stripper or
oxygen-generator system. In this case, the feed is at the top of the column
GAS TECHNOLOGY 89
and the pure is drawn off at the bottom of the stripper with reflux vapour
coming from the feed air boiling the in the sump. However, although
early plants were of this type, plants of this type are now unusual with
separation being based mostly on double-column units.
The Linde double column. The distillation of liquid air is frequently
required to provide two, three or more pure products (nitrogen, oxygen,
argon and other noble gases) simultaneously. The outline above would
indicate that for each product required at high purity another distillation
column is required. The rectification action normally provides a pure stream
of A and a stream depleted in A, rather than simply separating A from B and
providing two pure products.
It is possible, as in conventional petrochemical distillation, to use a feed
point in the middle of a column, simply using the top of the column as the
rectifier and the bottom as the stripper. This approach is used in a few plants,
for example in the BOC Groups Widnes UK plant, with a single column at
high pressure. However, these are exceptional plants where 100% liquid
and liquid are produced. In these plants the energy required to produce
liquefaction make any slight deficiencies in efficiency in separation insignifi-
cant. The necessity to operate a heat engine to provide a boiling and
condensing service is removed because that heat engine service is already
installed for the liquefaction capability.
The single-column and plants typically only recover 50 or 60% of
their air feed as product as they have considerable amounts of the wanted
feed-air component in the waste stream, which is the source of refrigeration
for the reflux they need. To achieve a high efficiency and a high recovery a
two-column system is required, where the waste stream from one column is
employed as feed to the other in some way.
By a happy coincidence it is possible, by choosing pressures correctly, to
employ the boiler in a middle-fed column (the main column) to cool a
nitrogen-column condenser and to use waste liquid from that higher
pressure nitrogen column to flash off and give reflux at the top of the main
column. The thermal linking of two columns operating different separations
at different pressures is a quite general opportunity in many chemical
separations although it is comparatively rarely used. The ability to condense
nitrogen with liquid oxygen depends on the boiling point of liquid oxygen at
5 or 6 bara being 3 lower than the boiling point of liquid nitrogen near
atmospheric pressure.
The advantages of this concept were first spotted and engineered by Carl
von Linde in his now classic double-column design (Figure 2.13). It has a
lower high-pressure rectifier section to which the feed air is taken, whose
condenser is the reboiler of the upper low-pressure main column.
The very common double-column design links the two columns also by
pipework as the feed air goes to the bottom rectifier and both liquid
90 INDUSTRIAL GASES
Figure 2.13 Linde double-column air-distillation unit.
nitrogen and rectifier bottoms are fed to the low-pressure upper main
column. The feed provides boiling service for the rectifier with condensing
service coming from the central reboiler/condenser. The main column feeds
the rectifier bottoms, which are of course an mixture, to its middle,
flashing them off to its lower pressure. For the main column, reboiling
service comes from the central reboiler/condenser and condensing service
from the flashed off liquid nitrogen fed to the top from the rectifier.
In principle, then, the Linde system provides pure and pure without
an external heat pump. However, practical designs are made more complex
by:
the need to deal with argon (which, of course, being heavier, ends up in
the stream;
the accumulation of uncondensible light gases, neon and helium, which

must be purged out of the top of the Linde rectifier column;
the need to provide waste streams which can be used to purge out
pre-purifiers or reversing heat exchangers;
the need for liquid product output, and addition of refrigeration cycles,
requiring a turbine at one or more of various possible points to carry out
expansion, rather than the inefficient JouleThomson expansion valves
indicated above;
complexities associated with the most efficient use of heat exchangers;
the need to avoid the accumulation of small quantities of hydrocarbons

GAS TECHNOLOGY 91
which may have escaped pre-purification and could end up mixed with
liquid oxygen and form an explosive.
Distillation column design. Sieve plates, consisting simply of a sheet with
many holes punched in it, are the mainstay of the cryogenic distillation
column. With shallow weirs (a few centimetres) sieve trays are made of thin
aluminium alloy sheet, often less than a millimetre. Vapour bubbles up
through the holes, frothing up the liquid on the tray, whilst liquid overflows
over a weir at the edge of the plate and flows down a wide pipe onto the plate
below, the weir height controlling the height of the liquid on the sieve plate.
Alternate plates have weirs on opposite sides of the column, thus forcing the
liquid to zig zag down. The liquid on each tray closely approachs equilibrium
with the vapour because of the intimate mixture of the two that occurs in the
froth which forms as the vapour bubbles through the liquid.
Distillation columns can have problems, however:
Weeping: liquid streams from a plate to that below effectively shorting

out the action of that plate. Weeping gives the lower limit to the column
vapour rate before a certain sieve plate design. Sieve plates with small
holes to give lower weeping will also have high gas pressure drop,
however, so a compromise design has to be adopted.
Foaming: liquid foam rises up and shorts out a plate to the plate above.
Foaming gives the upper limit to the column vapour rate. Foam in
distillation columns is not like the low-density stable soap/water foam
familiar in the kitchen: column foams are of much higher density, 20%
liquid or more, and are not stable, breaking back to liquid and gas in
seconds. They are nevertheless troublesome.
Sieve tray deflection: if gas flow is excessive, then trays will bulge upwards.
If design thickness is inadequate for liquid loading, then sagging will
occur. Also at high gas flows vibrations can be set up which can cause
excessive weeping or damage to the column in the long term. These
vibrations have been linked to synchronised bubbling of vapour through
the sieve holes, inter alia.
Channelling: liquid and vapour are not evenly distributed across the
diameter of the column. It is dealt with in packed columns by adding
distributor pipe assemblies every few pipe diameters. In plated columns
correct design of feed of liquid to the column at the top, the liquid path
from plate to plate, and strictly flat and horizontal plates usually avoid any
problems.
A measure of the efficiency of a column is the number of theoretical plates
in a certain size of column or the ratio between this and the actual number of
plates, i.e. the plate efficiency. The number of theoretical plates needed for
a particular separation is dictated by the composition diagram and the purity
of product desired but is generally quite small. The small losses of efficiency
92 INDUSTRIAL GASES
due to pressure drops bigger than necessary which are entailed by installing a
surfeit of plates are more than compensated for by the additional freedom of
operational range allowed by a longer column. In particular, it is often
convenient to operate columns with a relatively small reflux flow ratio,
which means that extra plates are needed.
Packed columns, generally structured packing comprising regular rows
of crimped foils rather than randomly packed columns of irregular mineral
or the manufactured Raschig rings, are an alternative to fabricated plate
columns. Height of equivalent theoretical plate (HETP) is the efficiency
measure used. HETP values are normally very small, a few centimetres even
with quite coarse packing, and this is why packed columns are widely
researched. Not only are packed columns smaller than trayed columns,
leading to lower capital costs, but there are potential running cost
requirements as well: lower cold box heat inleak losses and lower hydraulic
pumping losses due to the need to pump liquid up a tall column. Pressure
losses, due to having to pump vapour through the liquid and the energy
expended in creating the froth, are also avoided in packed columns, where
the essential gas/liquid contact is achieved by the vapour simply passing the
large area of thin film on the packing.
Packed columns are increasingly being used instead of trayed columns,
with companies like the cryogenic plant divisions of LAir Liquid and Linde
leading the way to plant entirely based on packed columns. They are not
invariably the correct solution, however, because problems with achieving
high turndown are common and packed columns generally have to be
operated within comparatively narrow limits of liquid and vapour flow rate.
In the opposite direction of performance is the bubble-cap tray. Bubble-cap
trayed columns have the advantage of high turndown and generally wide
operating limits, and high plate efficiency. However, in cryogenic service
they are generally regarded as having pressure losses which are too high and
large plate heights, i.e. much larger than the few inches necessary in trayed
columns.
A halfway-house between packed columns and trays in some respects is
the use of trays made of a porous material. These have been claimed to give
good gas flow while giving relative freedom from weeping and foaming.
However, no commercial plants are using this concept to date.
Finally, just becoming popular are refluxing exchangers. These are heat
exchangers designed for distillation service. The types being extensively
trialled now are mostly heat exchangers of the plate-fin type mounted in a
vertical plane which integrate the action of heat exchange and distillation.
Typically distillation takes place in vertical fin passages whilst horizontal
passages are connected to provide condensing and boiling service, and
maintain appropriate temperature distribution. Clearly, distillation based
on these units will have restricted regions of operational liquid and vapour
flow but within those limits they have the potential for very high efficiency.
GAS TECHNOLOGY 93
ASU detailed design. Once the basic cryogenic design has been complete,
there are a myriad of details to be determined on a practical air-separation
unit.
Instrumentation is a fundamental part of all ASUs, varying in complexity
from the humble thermocouple costing a few dollars to $50k or more for
multigas analysers such as process chromatographs. With all the cryogenic
part of the ASU hidden away deep inside its cold box, it is necessary to have
more instrumentation than a standard chemical plant. Levels at various
points in the equipment, as well as in the storage tanks, must be recorded
and the mainstay for level monitoring is the differential pressure (DP) gauge
of the type used in storage tanks (section 2.3.1). Several points going up the
column will be monitored for temperature, as will the different pressures in
columns and coming from compressors. Flow rate measurement of cryoge-
nic liquids is difficult to carry out accurately (section 2.3.1) and often
avoided. Gas flows, although also somewhat difficult to gauge accurately,
are useful and accurate venturi or turbine meter flow measurement is
necessary with a pipeline supply in order to be able to charge the user for the
gas used.
The compressor is a significant (and expensive) machine in its own right
and is normally monitored carefully for electrical input versus flow and
pressure output, giving approximate absolute efficiency and a good
measurement of any changes of efficiency. Parameters such as bearing oil
feed pressure and temperatures may be monitored in addition to the basic
input/output monitoring.
Early ASUs occasionally exploded when hydrocarbons accumulated to
dangerous levels in liquid oxygen somewhere in the system and the mixture
ignited. Although these incidents were rare and often of small scale, modern
plants take pains to avoid these problems by using additional absorber units
based on silica gel in key points in the design. These are occasionally
regenerated by being warmed up with a stream of purge gas. As hydro-
carbons accumulate in liquid oxygen, two absorbers are often used,
connected to the sumps of the lower column and upper column, absorbing
acetylene and methane, for example, from oxygen-rich and pure oxygen
liquids respectively. The gel absorbers work at cryogenic temperatures with
good efficiency.
Gas analytical instrumentation is essential both for monitoring the plant
status and also for checking product purity. For ASUs with argon
production, for example, it is useful to measure Ar content in the argon side
stream, which amounts to plant status information. A process gas chromato-
graph using zeolite molecular sieve columns is used. For a semiconductor
industry generator, output moisture and content are critical and
analysers much more sensitive than the norm will invariably be used to check
output. Occasionally, gas analysers can also have a safety aspect, for
example detecting the presence of traces of methane, acetylene and other
GAS TECHNOLOGY 95
hydrocarbons in the feed air or liquid oxygen may help avoid explosive
mixtures building up. Analysers employed for output monitoring can also be
used to give plant status with respect to some conditions.
10
Modern ASUs are fully instrumented in all but the smallest sizes with
numerous thermocouples, pressure, flow and level gauges and gas analysis
equipment as described. All the data from these sensors are led back,
typically via a number of programmable logic controllers and set-point PID
(proportional-integral-differential) controllers, to a central computer in the
control room of the site. This displays mimic diagrams of the whole plant,
indicating the gauge readings and positions of valves. Programs will be set up
on this computer for optimum performance in various modes. One typical
mode would be maximum liquid nitrogen production. During the night,
when electrical power is cheap, liquid nitrogen production is maximised.
During daylight, higher power cost may dictate a maximum gaseous oxygen
efficiency mode or even a minimum energy consumption stand-by mode.
At other times maximum argon may be demanded or a special nitrogen
purity may be required. Figure 2.14 shows a tonnage oxygen ASU with tank,
column in cold box and compressor.
Selection of distillation designs. In industrial practice there are many
different designs of cryogenic air separation units in use. Convergent design,
the mark of an efficient, maturing industry, is not always obvious in ASU
equipment. This is partly because the advantages of scaling-up plant are
such that standard off-the-shelf units are unattractive because the plant
needs to match quantities required precisely. However, the wide variety of
plants is also a function of the form and relative quantities of product gases
required. There are advantages to operating plant designed as modular
units, however, and this approach is seen occasionally in plants designed for
gaseous nitrogen which have been retrofitted with a separate bolt-on
cryogenic liquefier to produce liquid product.
Designs must be thought through if the correct decisions on plant design
and operating are to be made. If nitrogen is required at high pressure, 100
bar for example, then the most efficient plant may well be one operated at
(paradoxically) low pressure. The low-pressure product is compressed for
the downstream process, which means that large amounts of energy needed
for compression to 100 bar are not wasted on the waste streams of nitrogen
and oxygen.
Separating argon. Argon used to be regarded as a nuisance, contaminating
10
An example of the latter is the case of a plant where purity of the output liquid nitrogen was
being compromised by small amounts of By comparing the output analyser with results
from an intermediate analyser it was possible to show that the distillation process was fine but
that oxygen was somehow ending up in the product after distillation. Suspicion fell on a heat
exchanger which subsequently proved to contain minute cracks.
96 INDUSTRIAL GASES
Figure 2.15 Separation of argon using a side column.
oxygen and making it less effective in oxy-acetylene cutting, for example,
although its inertness means that it rarely caused any real trouble. With the
rise in uses for argon, however, from sputtering of coatings to welding of
100 mm steel plates for chemical plant vessels, argon is now separated out in
all cryogenic air-separation units of substantial size.
With its boiling point intermediate between liquid oxygen (LO) and liquid
nitrogen (LN), liquid argon (LAr) concentrations are highest somewhere
below the middle of the upper column in a double-column ASU when the
ASU is producing pure (99.5% or more) oxygen. If argon is not withdrawn,
it will accumulate in the distillation column until its concentration is such as
to contaminate the oxygen output. In plants producing low-purity oxygen,
argon is mostly lost in that oxygen and recovery may not be worthwhile.
In a typical ASU, a liquid product containing mainly oxygen, the
composition being typically 100 ppm LN, 10% LAr and 90% LO, can be
withdrawn. This can be further distilled in a separate column, which is often
placed alongside the main distillation column to avoid having to pump the
cryogens up or downhill and operated at a similar pressure to the main
column; this is the argon side column system (Figure 2.15). The side
column upgrades the argon content up to 95% or 98% or so Ar, the
produced argon being the top product. The bottom product, almost pure
oxygen, is fed back to the main column a little further down below the
take-off point. The condensing service for the column is provided by
expanding a small amount of liquid air from the bottom high-pressure
column across a valve and passing the cool air through a condensing heat
GAS TECHNOLOGY 97
exchanger at the top of the side column. The expanded air is wasted via a
heat exchanger or fed back into the low-pressure column nearer to the
nitrogen end.
The industrial or crude argon produced by the side column is saleable to
some users, for example those welding shield gases where some oxygen is
allowed in the inert argon mix. For many crude argon applications low
nitrogen content is more important than the few per cent residual oxygen.
For most users, however, a few per cent oxygen is completely unacceptable
and a true inert gas is needed with only parts per million of reactive gases
such as oxygen or even nitrogen. This pure argon is often produced by a
standard sequence of a catalytic reaction with excess hydrogen to remove
oxygen, followed by an additional distillation step.
The distillation step is preceded by a condenser and a temperature-swing
drier, where water formed in the catalytic step by hydrogen/oxygen
combination is removed. In the additional stand-alone distillation column
(it is little more than a flash-off tank), excess hydrogen and residual nitrogen
are removed as lights and liquid argon removed as bottoms. The catalytic
reactor, because of the high degree of purity of the gases used, can be of the
precious metal/inert support type without fear of loss of activity from
poisoning.
Separating neon, krypton, and xenon. The inert gases He, Ne, Kr and Xe
are exceptionally difficult to detect in the air and their separation from it is
no less difficult.
Helium and neon are left dissolved in liquid nitrogen separated in a Linde
double column-system. By expanding the liquid in a phase separator prior to
storage, a top gas can be obtained with several per cent neon and helium
together. ( Helium is inexpensively obtained by processing of natural gas
(section 2.2.8), so neon is the main product of the system.) However,
commercial production of neon is carried out in a minority of large
air-separation plants by drawing a neon-enriched vapour stream from the
top of the bottom (high-pressure) column of the Linde double-column unit.
The vapour is taken to a neon side column where it provides boil-up vapour,
the condensing service being a liquid-nitrogen-fed heat exchanger con
denser at the top. The crude neon can be pulled off from the top of this
column at 50% efficiency and at 50% concentration, the balance being
nitrogen and helium with minor amounts of hydrogen.
Neon is further purified by adsorption onto porous beds, typically
activated carbon. First, hydrogen can be removed if necessary by catalytic
combustion. Next come carbon adsorption units. The final bed in a
multi-stage, multi-bed system is often a LN-cooled carbon adsorption
temperature-swing regenerated unit, while the first stage is usually a
pressure-swing regenerated unit at room temperature.
The final stage of purifying neon is to liquefy it either by a mechanical
98 INDUSTRIAL GASES
expansion cycle or by cooling with a liquid hydrogen or gaseous helium
cooling circuit. After liquefaction, remaining helium is lost by flashing off
the liquid neon to atmospheric pressure, recycling the top-gas. Neon is now
supplied in relatively large quantities. It is packaged both in compressed
cylinders, on tube-trailers of cylinders and in liquid form in small cryogenic
dewars, almost identical to those used for LN. (Neon production was greatly
expanded in the USA for a while in the 1980s as a result of US Department of
Defense needs.)
Krypton and xenon are only present to the extent of 1 and 0.08 ppm in the
air, making them exceptionally difficult to separate. In practice separation is
carried out in several steps:
1. Withdraw an LO stream: this, in principle, will have Kr/Xe at 5 and
0.4 ppm.
2. Boil up LO in a distillation side column, use N in the condenser,
producing a concentrate containing 0.2% Kr and 0.02% Xe in oxygen.
3. Burn-out residual hydrocarbons in a catalytic combuster (the excess
oxygen providing oxidation service).
4. Absorb
5. Distil cryogenically again, the bottoms containing up to 80% Kr and 6%
Xe (oxygen being a top product).
6. Burn-out residual hydrocarbons again.
7. Absorb CO
2
/H
2
Oagain.
8. Bottle the krypton/xenon mixture for sale or accumulate enough Kr/Xe
to liquefy against LN and distil to pure Kr and Xe.
9. Bottle Kr and Xe.
2.1.6 Pressure-swing absorption
PSA adsorbers and molecular sieves. It is usual in chemical engineering to
distinguish between absorption processes, where a gas is absorbed by being
dissolved in the bulk of a liquid, and adsorption processes, where a gas is
adsorbed in the surface of microporous solid. A biological example of the
former is the absorption of oxygen by haemoglobin in blood whilst in
industry liquid absorption is used to extract carbon dioxide and hydrogen
sulphide from gas streams. Absorption of acid fume and toxic reactive gases
by water spray scrubbers is another example. Adsorption by solids is met
most commonly in the form of activated carbon granules, which are used to
remove impurities in air in many applications. In the latter two cases, the
process is often not reversed; the water and activated carbon are simply
dumped when their capacity is exhausted. For separation of gases, as
opposed to removal of trace impurities, a cyclic process with regeneration of
the adsorber is necessary. Pressure-swing adsorption (PSA) is an example of
such a reversible adsorption process.
GAS TECHNOLOGY 99
PSA units for air or gas separation mostly rely on molecular sieves.
These are not sheets of material with very small holes through which a gas is
forced leaving the big molecules behind (this is the principle of membrane
separation described in section 2.1.7). They are in fact microporous solids,
with molecular-size pores, which can be used to adsorb and desorb gases.
Mostly because of the different forces of attraction between different gas
molecules and the micropores, but also because of the different size of gas
molecules, the adsorption differs between gases and can be used as the basis
for separation.
Almost all microporous materials will show separation effects, at least for
some mixtures of gases, but few satisfy more stringent criteria for a useful
absorbent:
reasonable crush strength so that they can be compacted into tall vertical
beds without blocking;
little tendency to dusting, i.e. converting from granular solids into fine
dust;
controllable microporosity so homogeneous beds of known capacity can

be prepared;
not too strong a binding to any one gas because desorption will then be
difficult;
small macropore/interpellet void volume: materials with a large volume

which is not in the active micropores require larger purge volumes to
achieve high purity products and high recovery;
macropore structure which allows good access for gas flow through the
bed and to the micropores.
These criteria have all but eliminated most natural products and many less
reproducible synthetics such as calcined clays. Early pressure-swing adsorp-
tion was based around activated carbon or the porous solids well-known as
easily-regenerable drying agents. Silica gel, made by neutralising strong
solutions of sodium silicate acid (water glass) with acid, and hydrated
alumina, made by dehydrating hydrated alumina, are examples of the latter.
Silica gel and activated alumina both retain water molecules strongly, which
makes them good materials for use in a gas drier unit but less useful in gas
separation where the gas mixture may also contain moisture. However, the
moisture in these materials is loosely bound and can be removed by heating
to 150C or by pressure reduction as part of a PSA cycle. Hence the first part
of the gas-separation bed using alumina or silica gel in fact functions as a
built-in drier unit.
Activated carbon has been a big industry since the extensive nineteenth
century investigations into the many interesting properties of carbon.
Carefully prepared pellets of carbon with volatile binders are first heated to
500C or so to eliminate the binder and other hydrogenous material,
followed by stronger heating in special atmosphere, typically containing
100 INDUSTRIAL GASES
mild oxidants such as or steam, causing the slow development of
microporosity throughout the bulk of the material. Activated carbon is
particularly useful in separating gases which may contain moisture because
its surface is almost non-polar and will not attract and retain water
molecules.
Two types of microporous solids are now of importance in gas separation
by pressure swing: carbon molecular sieves and zeolite (aluminosilicate)
molecular sieves. Molecular sieve materials are the only materials capable
of separating air gases: even on typical molecular sieves the ratio of to
adsorption is only 2 or 4 (with alumina or ordinary activated carbon the
ratio is close to 1).
Molecular sieve carbon (MSC) was developed by applying activated
carbon treatments to particular carbon sources or by coating the pores of
conventionally made activated carbon with more carbon. One method uses
a thermosetting polymer which is then cooked on to form an additional
layer of carbon or carbonised material. Heating MSC materials in hydro-
carbon gases can deposit carbon in the micropores, adjusting the pore size.
MSC has, in general, more micropores within the active size range of
importance in gas separation (5-15 ) than the microporous solids above,
which have little porosity below 20 . Zeolites also have microporosity
below 10 , which is intrinsic to the crystallography of aluminosilicates.
The synthetic zeolites used are made by crystallising a sodium/aluminate/
silicate water gel by heating under pressure. The resulting crystals are in
effect clathrates: a continuous structure with cages containing water
molecules. On calcining above 500C, the water is eliminated to yield the
zeolite. Zeolites absorb water strongly and this can only be removed
completely by heating to 300C or so, meaning that gas mixtures to be
separated should in general be dried before an adsorbtion separation is
attempted.
MSC and zeolite both have structure at several levels: macropores which
allow bulk flow of the input mixture and purge gases, pores which are
microscopic but still large compared to gas molecules, and finally the
micropores, which are the part of the structure that shows appreciable
differences in absorption of gases. In the case of carbon molecular sieves
there is a continuous spectrum of pore size whilst in zeolites there is a
bimodal distribution with micropores of a few A and macropores between
crystallites of a few microns.
Some molecular sieves have been doped with a chemical agent to make
them more selective; an example is the BP carbon monoxide adsorber which
adds copper ions to a zeolite molecular sieve. In others the proportion of
cations in the zeolite aluminosilicate lattice of a zeolite is modified to change
the pore size and substitution of sodium and potassium atoms yields the now
standard 4, 5 and 13 pore sizes. PRAXAIR patents also name lithium as a
useful cation substitution in its zeolites. (PRAXAIR, formerly the Union
GAS TECHNOLOGY 101
Carbide gases business, carried out much of the early work on synthetic
zeolite molecular sieves and developed the market for the basic 4 , 5 ,
13 , etc., sieve types.)
PSA/VSA concepts. Pressure-swing adsorption is a simple concept in
which gas is admitted to an adsorber bed at high pressure then desorbed at a
lower pressure. There is either a difference in adsorption rate or a difference
in static amount of adsorption. Hence the two gases present in the input gas
are differentially adsorbed; one is concentrated in the bed whilst the other is
released in the depressurisation stream. After the depressurisation stream
has been flowing for a preset time to the outlet, the depressurisation is diver-
ted to waste until it is at near-atmospheric pressure, releasing the adsorbate,
perhaps with the aid of a purge stream.
In a so-called vacuum-swing adsorption (VSA) unit, the depressurisation
is continued past atmospheric pressure. In either case, this part of the cycle
ensures that a minimum of the adsorbed gas is left so that the gas mixture is
not distorted when the next cycle of raw gas mixture admission begins. The
trade off is, of course, the cost of installing and running the necessary
vacuum pump. The size of vacuum pump needed climbs very quickly with
required throughput since vacuum pumps have to handle large volumes to
achieve reasonable mass flow rates. However, the use of the vacuum swing
enables more efficient use of molecular sieve. Typical absorption isotherms
do not favour PSA operation with a high delivery pressure so most PSAs are
best operated with venting down to near 1 bara, followed by a compressor to
step up pressure if necessary. Vacuum-swing absorbtion units for oxygen
have shown greater economy than PSAs even when the extra cost of buying
a large vacuum pump is taken into account. The vacuum-swing cycle makes
more effective use of the isotherm and power efficiencies can be increased by
75%. VSA also usually allows a simple, low-cost two-bed design where a
PSA system might require multiple beds.
More exotic pressure-swing systems are continually being proposed. Sys-
tems with large numbers of beds, all undergoing different processes at dif-
ferent pressures simultaneously, are typical. One ideal system to which
some have aspired is a moving bed system in which gas coming in is continu-
ously met by fresh absorber, which is slowly removed and regenerated in
such a way that the whole system operates in a dynamic steady state. Systems
have been proposed which simulate the pseudo-moving bed SORBEX
organic solvent separation process. SORBEX swaps process streams be-
tween dozens of inlet pipes to a single large fixed adsorber column to simu-
late the moving bed. One more recent exotic system proposed is pressure
parametric pumping (PPP). PPP involves pumping gas through a column in
one direction at high pressure and in the reverse direction at low pressure,
using a specially designed pump. The high-pressure outlet end concentrates
the less absorbed component even when the cycle time, the pump rotation
time, is a few seconds because the gases migrate down the column a little
with each pump cycle.
Isotherms and adsorber bed dynamics. In order to use the adsorber beds
and the compressor energy efficiently, it is necessary to make the depressu-
risation to product stream as long as possible; the longer this goes on, the
greater the output of desired gas for a given plant and energy consumption.
However, the longer the desorption is carried out to the product stream, the
more the product will be polluted by the other elements of the mixture
breaking through.
There are some tricks which can improve matters, however. Absorber
bed materials which have a favourable isotherm will have a sharp
breakthrough characteristic. This means that nearly all the bed capacity can
be exhausted before breakthrough occurs if flow patterns through to the
active micropores permit. Nearly all practical bed materials have this
characteristic.
The isotherm of an absorber is the graph showing the amount, Q, of a gas
absorbed at equilibrium versus concentration, C, of that gas at a particular
temperature. A favourable isotherm is one which does not show a linear
relation of Q to C but shows a shape closer to that of an irreversible reaction,
i.e. Q grows rapidly at small C but then saturates at higher values of C. Such
an isotherm will tend to sharpen the concentration front passing through an
absorber bed, as high-concentration fronts move more rapidly than
low-concentration fronts, so that an initial sloping front will tend to grow
into a very sharp shock wave and approach a step change in concentration.
A commonly used isotherm is that named after Langmuir. Langmuir
based his simple theory on an equilibrium between adsorption and
desorption in the following way. Kinetic theory gives as
the collision frequency of a gas. Rate of desorption is a thermally activated
process with an activation energy given by the heat of adsorption ( H) of the
gas. Hence rate of desorption is proportional (with constant b) to fractional
coverage of adsorbate (f) on the material surface multiplied by a Boltzmann
factor Thus, if a is the sticking probability of a gas molecule hitting
the surface
or
where
B = Langmuir constant =
and m is the molecular mass. Figure 2.16 plots Langmuir isotherms for a
typical molecular sieve absorber at two different temperatures.
GAS TECHNOLOGY 103
Figure 2.16 Langmuir isotherm plots.
With an isotherm of this shape, clearly higher concentrations will be less
absorbed relative to lower concentrations, as required to sharpen the
concentration front. In practice, absorber beds do not show the sharp fronts
expected from this simple theory because
flow into the micropores of the absorber is not instantaneous;
absorbers are prepared in rather large pellets, with large voids between
them, to avoid pressure drop;
the bed will tend to heat up as adsorption takes place;
zeolites have two levels of microporosity;
pellets will not in general be identical spheres (the usual assumption in

modelling).
One conceptual model, which approaches more nearly the practical
situation, considers the absorber bed to be a set of well-mixed tanks
connected by pipes. In this way a simple numerical model can be built which
can be tuned to match empirical measurements in the laboratory or on small
units, the model being then usable to help in the design of full-scale units.
Many absorber beds in practice show S-shaped concentration fronts
(Figure 2.17). These quite typically show some broadening as the front
progresses through the bed, despite the sharpening effect of the typical
favourable isotherm. In long beds, however, a constant MTZ (mass transfer
zone) develops, representing an equilibrium between the front-broadening
due to mass transfer within the pellets of adsorber and the front-sharpening
due to the isotherm. In this case, a constant S-shaped curve is seen
progressing through the bed. Often an effective MTZ is assumed in order to
carry out approximate calculations.
Figure 2.17 Concentration front curves progressing through a PSA bed.
PSA design considerations. A PSA must regenerate its adsorber material,
hence, in addition to knowing when impurities will break through an
adsorber bed, it is also necessary to know when desorption of those
impurities by vacuum, purge gas pressurisation or flow is complete. An
essential part of the economics of PSA is having a sufficiently short
regeneration time for the adsorber bed whilst not wasting excessive amounts
of compressor energy or feed gas.
The heat of adsorption means that a PSA bed is not isothermal. A typical
adsorber may rise in temperature by 100C or more during adsorbtion and
fall by a similar amount during desorbtion. (It has even been proposed to use
the cooling effect in a PSA unit as a kind of refrigerator; one having the
advantage of using no CFCs, of course.) Also, the compressed supply may
be not cooled completely to room temperature. To a large extent, the
non-isothermal nature of the bed can be controlled simply by increasing the
size of the beds relative to their total possible capacity or, equivalently,
choosing a much shorter cycle time than would be indicated. This is the
approach generally taken.
Other approaches to non-isothermality are possible, however. Relatively
small heaters installed near the feed end of beds help as this part of the bed
tends to cool below ambient, limiting desorption efficiency during depressu-
risation. Another approach is to mix the adsorbent with an inert substance
with a high heat capacity. Candidates are those which are conveniently
stable and inexpensive, and have not too low a thermal conductivity. Iron
has been used but non-microporous alumina, magnesia or aluminium would
appear to be candidates, with alumina and magnesia offering safety
advantages over iron with oxygen-containing gases, and aluminum offering
better thermal conductivity.
GAS TECHNOLOGY 105
Uneven distribution of flow can seriously degrade a PSAs bed concen-
tration profile, even in the presence of a favourable isotherm. The packing
of the bed must be done very carefully not only so that it gives even flow but
also so that it does not move, even during the pulsed gas flows intrinsic to
PSA operation. Interpellet voids are unavoidable; they result in a sizeable
dead-space volume with a typical absorber. It has occasionally been
proposed to include, along with the large pellets, smaller pellets to fill in
between the larger pellets. With careful control of the relative numbers of
larger and smaller ones, it would be possible to decrease dead-volume
substantially while still leaving good flow characteristics. (It is well known in
road construction that a combination of coarse and fine aggregate roadstone
will settle to a much higher density than either on their own whilst still
maintaining porosity.)
Additional small purge flows after completion of depressurisation help to
remove residual impurities and improve purity in PSA units. The slight loss
of product recovery is more than compensated for by the increased purity of
product.
The PSA unit as described needs an intermittent supply from the
compressor and itself provides an intermittent stream of product. In piston
engines the intermittency of output can be dealt with by linking several
cylinders, which provide a sequence of closely spaced power pulses, and by a
flywheel to store some energy from one power pulse until the next pulse.
PSA units use both these approaches. They may employ an air receiver
vessel to store compressed gas from the compressor at the inlet and a
product receiver vessel to store product. Almost all practical units use at
least two adsorber vessels; one which is providing product whilst the other(s)
is being repressurised or purged. The two-bed system can be regarded the
mainstay of PSA plant and the standard by which other plants are judged
(Figure 2.18).
The sequence of operations in a PSA plant is an important part of its
optimisation. Despite the apparent complexity of many large PSA se-
quences, however, there are really only four essential steps going on:
1. pressurisation/production (on-line bed)
2. vent
3. purge to lower partial pressure of adsorbed species at 1 bara total
pressure (or use vacuum)
4. backfill to push remaining unwanted gas out
(5. pressure equalisation as an energy conservation measure makes five
steps).
Large improvements in efficiency can be obtained for gas separations by
increasing the number of beds and suitable switching arrangements between
them.
With more beds there is less variation in load on the compressor and less
Figure 2.18 Simplified two-bed pressure swing absorber air separator, (a) (b)
variation in product supply pressure, obviating the need for surge vessels.
There is also the possibility of including multiple pressure equalisation steps
in the PS A cycle. Pressure equalisation economises on compressor power by
raising and lowering bed pressure in steps by connecting beds to each other
instead of only to the inlet, waste and outlet streams. In a four-bed system
(Table 2.5) each bed is both lowered and raised in pressure in two steps.
Both energy requirements and product recovery are enhanced relative to an
equivalent unit with two beds. The steps 1 to 9 are of different lengths and
GAS TECHNOLOGY 107
are phased by microprocessor sequence controllers so that one bed is always
on line whilst the other beds are purging or equalising.
Hydrogen PSA. Hydrogen extraction was the first major process, after the
now ubiquitous PSA drier (considered in section 2.2.11), to switch to using a
PSA process to a significant extent. Hydrogen is particularly easy to separate
using PSA because of its low molecular weight. The hydrocarbon gases and
CO
x
mixtures it is found in are all easily absorbed by molecular sieves,
leaving hydrogen in the gas stream. Because hydrogen is typically extracted
in very large units (e.g. steam reformer plants, catalytic crackers in oil
refineries and ethylene plant by-product streams which are all often in the
range of 1000s of tonnes per day), these are large and sophisticated
multi-bed units.
The PRAXAIR hydrogen PSA process uses beds packed in the first
section with activated carbon for dealing with the highly absorbed species
(carbon dioxide and water) and the second section with 5 zeolite for
dealing with the less highly absorbed species (such as methane and carbon
monoxide). Product purities can exceed 99.999% on 70% hydrogen
feedstock and units with a million cubic metres per day (100 tonnes per day)
capacity are operating.
Nitrogen and dynamic adsorption PSA. Nitrogen PSA separation relies on
carbon molecular sieves whose adsorption rate for nitrogen is different from
that for oxygen. It is important to note that carbon molecular sieves have the
same equilibrium absorption properties for nitrogen as they do for oxygen.
In order to separate efficiently nitrogen PSAs operate quickly relative to the
gas absorption rate and never approach the equilibrium state where their
selectivity factor reduces back to zero.
A typical PSA plant of a few tens or hundreds of cubic metres per hour
employs an oil-inject screw compressor as its motive power. The use of
oil-injected screw compressors and other wet compressor types carries
some risk with it because oil can be carried over into the PSA carbon bed
where it soon poisons the adsorber. However, these compressors are most
efficient in smaller sizes; precautions such as additional preventative
maintenance measures, additional oil traps and cut-out devices are there-
fore indicated.
Small-scale generation, a few litres per minute of nitrogen, is possible with
PSA, although no such systems have yet become at all common. One
application which is a candidate for such micro-PSA systems is the
production of nitrogen for the dispensing of beer (section 3.4.3). Another is
the production of a small, steady stream of nitrogen for gas chromatography
(particularly valuable for GCs which are used for continuous monitoring
since it avoids tedious cylinder changing). By accepting a very low yield
relative to the input compressed air, generous sizing of beds (relative to the
very low flow rate of less than a litre per minute) and other optimisation, it is
possible to produce a 1 ppm oxygen level nitrogen.
Oxygen and equilibrium adsorption PSA and VSA. Oxygen PSA relies on
artificial zeolite molecular sieve materials which have, at equilibrium, a
difference in their retention of oxygen and nitrogen. Oxygen PSA systems
have a cycle time which is long relative to the gas absorption time constant,
so that equilibrium is almost reached, maximising capacity and relying on
the fact that their selectivity at equilibrium is still good. As mentioned
above, oxygen systems are often simplified and rendered more efficient by
using a vacuum pump to vent the beds after adsorption. These systems are
referred to as VSAs.
Compressed air for use in a zeolite PSA must be reasonably dry, as all
zeolites will preferentially adsorb until saturated and fail to function as
oxygen separators. Fortunately, the compressor will in any case intrinsically
remove the bulk of the water (see footnote in section 2.1.5) but further
removal is still needed. Zeolites are, in fact, themselves frequently used in
gas drying systems because of their strong but reversible adsorption of
water. Zeolites are best heated to remove moisture effectively and heated
bed driers (temperature swing absorption, TSA) are possible. The problem
is solved if a twin-bed PSA drier, perhaps using silica gel or alumina, can be
used in front of the oxygen PSA. A more common solution is to use drier
materials such as activated alumina in the first part of the adsorber bed. The
activated alumina has a more easily reversible binding to the water
molecules than zeolite and therefore water is more easily desorbed during
the blowdown and purge steps.
Many tens of thousands of oxygen PSA systems have been sold worldwide
and probably 99% of these have been small systems for one special
application; supplying a few litres per minutes of oxygen-enriched air for
medical purposes. These systems are about the size of a small suitcase and
require a power supply of 500 W or so, making them a thousand times
smaller than a typical industrial oxygen system. Tiny systems such as this
GAS TECHNOLOGY 109
Figure 2.19 Small-scale PSA oxygen generator for medical use. The unit hinges apart as shown
with, on the left, four large al umi ni um canisters containing molecular sieve adsorber and, on
the right, the compressor, gas analyser and control units. Pneumatically driven poppet valves
(five are visible on the left at the top) are used (courtesy of Puritan-Bennett).
must minimise complexity and hence cost; the whole system is typically
$2000 or so. A very simple implementation of the straightforward two-bed
PSA cycle is used in many units, with two beds of zeolite protected by silica
gel dessicant. Their performance is relatively poor compared to industrial
units, with only 30% or so recovery, for example. A difficulty is the
compressor, which must run continuously and very quietly, and yet supply
100 standard litres per minute of air. To minimise compressor cost and
noise, low pressures are used, less than 2 barg in some designs, and the
compressor is housed in an elaborately padded box. The purity achieved
varies from only 50% to 95% or so, depending on the design, but even the
lower figure is entirely adequate for the purpose. Figure 2.19 shows a typical
medical oxygen concentrator.
As mentioned above, the most power-efficient oxygen systems employ a
vacuum cycle purge. Compression to a low pressure, less than 1 barg and
sometimes as low as 0.25 barg, is used to produce oxygen from a freshly
regenerated bed at just above atmospheric pressure. This low pressure feed
is acceptable only because purging/desorbtion will be at vacuum pressures.
The gas is then compressed if necessary. This is efficient as only the product
(20% of the overall system gas flow) is compressed. The vacuum cycle
removes nitrogen more efficiently than an atmospheric purge cycle (section
2.5.10) because of the large ratio between absorbing pressure and desorbing
pressure, and because it is a better match to typical zeolite isotherms.
Two-bed systems are common in VSA designs, despite their relatively large
scale (up to 100 tons/day or so), since the improvements in power efficiency
do not justify the additional capital expenditure on a more complex system.
To minimise the complexity and hence cost of very small systems, the
Union Carbide Corporation, now PRAXAIR, developed earlier work
which used a single bed. With surge vessels smoothing the varying load on
the compressor and variations in the supply rate, an inexpensive system can
be devised. The size of the whole system, bed and surge vessels, is minimised
by using the fastest possible cycle time, i.e. just a few seconds. Gas
throughput is very high for the unit size. The valve layout with one bed is
very simple, with just two valves on the inlet side of the bed governing
whether the bed is pressurising or depressurising/backpurging to waste, and
the outlet surge vessel accumulating the less highly absorbed oxygen. The
single bed must use a finer grade of pelleted zeolite; the 1 to 3mm pellets
normally used are too large. The fine pellets mean that the bed has a high
pressure drop across it, which is required for the system to work. The
single-bed system would appear to be well-matched to applications such as
medical oxygen generation.
Practical PSA considerations. The control of a PSA unit can be as simple as
a timer which switches a spool valve from one position to the other at fixed
intervals. The use of a single complex spool valve appears to have
advantages in PSA designs because such a single valve can provide all the gas
switching with, as well, low-pressure drops. For many high-production rate,
low-purity plants, this may well be the correct choice. However, spool valves
are probably less reliable in general than poppet valves. Spool valves have
large numbers of sliding seals and are made in small numbers to special
designs; the converse of poppet valves. Leaking valves are only important,
of course, where the purity of the product is important but purity is often
important for nitrogen units. The use of poppet valves also enables the
maximum advantage to be gained from modern programmable logic
controllers. A PLC can drive many valves at different timings, optimising
the PSA efficiency, and can even be used to change timings dependent on
product stream needs.
PSA beds are intrinsically subject to large transient flows of gas as beds
switch, although the multibed systems with one or more pressure equalis-
ation steps are superior. During these brief pulses of flow, the adsorber
materials can move slightly. Once this starts to happen, the pellets tend to
abrade and generate fines, and become loose so that they can abrade more.
GAS TECHNOLOGY 111
The gas inrush causes fluidisation eventually, the whole bed starts to
abrade and, in the long term, to break down. Furthermore, channels may
form in the bed causing inefficient distribution of feed gas to the adsorber.
To restrain these undesirable effects sieve materials must be carefully
chosen for good crush strength and packed with some compressive stress
built in.
There is a trade-off between purity and production rate in PSA systems.
If some of the adsorbate is discarded, then purity can be improved. How-
ever, as adsorber materials improve, then the trade-off shifts to higher
purity or higher efficiencies. It is important to recognise that economically
effective PSA systems are only currently useful for relatively low-purity
nitrogen and oxygen. (Although, as noted in section 2.4, they can achieve
much higher levels of purity on hydrogen, and low yield systems may
achieve high purity.) Oxygen of purity better than about 95% or nitrogen
of greater than about 99.9% are currently probably best derived from de-
livered cryogenic liquids. Alternatively, the additional costs and com-
plexity of a downstream gas purifier unit must be accepted.
Today, much computer modelling can be done to guide PSA design and
operation; both at the micro-scale of absorbers and gas passages within
them and at the macro-scale of sequencing of valve operation and control-
ling flow rates during the sequence. The importance of modelling is that it
allows the scaling of plant; custom-made PSAs for nitrogen and oxygen are
now being made at the 100200 tonnes per day level and this maximum size
has grown by a factor of two every four or five years. PSA units for lower
purity gases from air separation are up to twice as efficient in energy terms
as cryogenic air separation, so there is a strong driving force behind this
size growth. The early role of sieve supplier companies in computer model-
ling and experimental work, like Bergbau Forschung in Germany, will
most likely be overtaken by the gas companies as PSA grows to be more
and more a core technology for the latter.
2.1.7 Membrane separation of gases
The membrane is the simplest conceivable device for gas separation. The
ubiquity of filters to remove dust particles from gas leads naturally to the
assumption that a similar separation can be achieved for gas molecules
simply by decreasing the size of the filter holes. However, it turns out that
using filter membranes for gas separation is much more difficult. Gas mol-
ecules are only fractionally larger than the atoms of which the membranes
must be made. Sieving membrane filters are impossible because of the im-
possible thinness of the membrane required, so membranes can at best act
as deep filters. However, one gas molecule is very much the same size as
another, whereas a dust particle of 0.1 diameter is hundreds of millions
Figure 2.20 Cascaded gas separation membranes.
of times as big as the gas from which it is filtered, so filtration mechanisms are
quite different.
Uranium and membranes. Membranes have a long history of experimen-
tation but very l i t t l e history of actual practical application. One of the most
famous examples of separation, and the first major application for mem-
branes in gas separation, was the giant Manhattan Project of the Second
World War. This employed dynamic separation across semipermeable
membranes to prepare one isotope of uranium, for the first atomic
bomb. The separation factor was very small indeed; just that due to the
difference in speed of the two uranium hexafluoride molecules, which is
proportional to the square root of the ratio of their masses, 1.00429.
To obtain useful separation, and to avoid throwing away most of the
a cascade of membranes was used, with the permeate and retentate of the
first unit being led away to two subsequent units whose permeate and
retentate were in turn led away (Figure 2.20). Even so, the separation
achieved was highly incomplete.
Fortunately, only partial separation of the isotopes was needed and a
large amount of money was available for the job. The resulting plant cost
billions of dollars in todays money, covered square miles and required
hundreds of megawatts to run it. A modern successor to the Manhattan
Project is the French plant at Pierrelatte. This leviathon amongst chemical
plants actually requires the power of four nuclear power stations to run it,
GAS TECHNOLOGY 113
whilst each of the pumps propelling the through each membrane is of
megawatt size.
Nitrogen from membranes. Fortunately, the production of nitrogen via a
membrane method is considerably easier than the separation of uranium.
Rather than using a semipermeable membrane, air separation uses com-
pletely sealed but very t hi n polymer membranes which are much more
efficient. However, nitrogen has a very low market price compared to
enriched uranium and it is only very recently that economic plants have
become available.
The breakthrough that has enabled cost-effective nitrogen production is
the advent of chemically selective membranes. Rather than relying on the
small difference in molecular weight or other physical properties, as the
uranium separation does, selective membranes achieve orders of magnitude
better separation in a single stage by use of a chemically active membrane.
A semipermeable membrane relies on microporous structure. However,
the micropores are in fact much larger than the gas molecules being
separated. They simply ensure that no bulk flow of gas takes place through
the membrane. Then the transit of gas across the membrane simply depends
on molecular speed. An active membrane works by allowing the gas to
dissolve, diffusion then causing the gas to migrate through it. Separation is
both by means of the differential solubility of the gases and by their different
diffusivities.
Gas flow, rate F, across a membrane is given by Ficks law
where is the membrane permeability, t its thickness, A its area and the
initial and final pressures. Hence membranes of the maximum area and
smallest thickness are needed, whilst large pressure drops are inevitable.
A typical membrane separator has a thin (1000 ) active but not very
permeable membrane supported on an easily permeable non-selective
(microporous) membrane. This kind of structure is known as an asymmetric
membrane and was first developed in the 1960s using dense (selective)
cellulose acetate on low-density microporous cellulose acetate (non-
selective). The asymmetric structure enables the use of the thinnest possible
active membrane and enables the use of mechanically weak selective
membrane polymer. Membrane areas are claimed for fibre modules which
have 10000 of membrane for each cubic metre of module volume.
Membrane chemistry is shrouded in commercial secrecy but many mem-
branes are simply polymers such as unplasticised PVC, cellulose acetate,
polysulphone or silicone. A few results are now published on metalorganic
complex compounds. Polysulphone and acetate membranes give oxygen
permeabilities four or five times that of and this is typical of the simpler
commercial membranes, although these have greater permeabilities for
both components. Permeabilities increase with temperature according to an
Arrhenius activation relation, with activation energies in the range of 1 to
3 kJ
where is permeability, is permeability at a reference temperature,
is the activation energy for the process, k is Boltzmanns constant and T is
absolute temperature.
There are thus considerable increases in permeabilities due to tempera-
ture and increasing temperature of operation from 20 to 65C, for example,
would give, with an activation energy of 40 kJ a ten times greater
throughput. Typical relative permeabilities for plastics are
10 0.2 1 (reference) 0.2 6
Within these relative permeabilities, which vary surprisingly little for most
plastics, there is a millionfold variation in absolute value of permeability;
specially made barrier films, such as polyvinylidene chloride, have
permeabilities which are 1/2000 times that of polyethylene polystyrene or
Teflon, whilst dimethyl silicone rubber has a permeability which is a
hundred or so times greater than these. Silicone and natural rubber are
anomalous in that they do not show a narrower range of relative
permeabilities than the ratios shown above. There is an unhelpful trend that
high relative permeability ratios are seen only in low permeability polymers.
Hydrogen, for example, permeates only a factor of two faster than oxygen
on rubber. Some oxygen permeabilities for commercial polymer films (in
are:
Rubber Low density High density Nylon 6,6 Poly- Polyacrylo-
polyethylene polyethylene vinylidene nitrile
(LDPE) (HDPE) chloride (PAN)
(PVDC)
18000 1900 430 30 5 0.16
Commercially, nylon, polyvinylidene dichloride (PVDC) and polyacrylo-
nitrile (PAN) are regarded as barrier polymers. Permeabilities are some-
times measured in units of at standard
temperature and pressure. This conveniently gives oxygen permeabilities a
number little bigger than unity for the common plastics and only a factor of
few hundred for silicones. The presence of the factor is a reminder that
permeation is not an easy process.
These figures also have relevance for the design of many other artefacts.
In food packaging (section 3.4.3), it is clear that carbon dioxide will be lost
quickly from packaging whilst nitrogen can largely be retained. For food
GAS TECHNOLOGY 115
packaging membranes permeability of oxygen is often quoted and labelled
oxtran (oxygen transmission).
Membrane separation is highly effective for the separation of heavier
organic or toxic components from air; a fact that is occasionally employed in
instruments for the monitoring of toxic gases e.g. mass spectrometric
analysis on molecules passing through a silicone membrane. One such
instrument (VG Instruments, Middlewich, UK, model PETRA) claims
thousandfold or more enhancements of toxic trace gas component peaks
using the technique.
The figures also make it clear that membrane separation of from air or
methane is straightforward. This separation was the subject of the first
commercially successful selective membrane process, the Monsanto PRISM
process, based on a polysulphone selective membrane. The PRISM
membrane is an asymmetric membrane moreover, although a crude
separation could be achieved using a simple membrane. Monsanto mem-
brane units are already in service carrying out reclaim from ammonia
plant gases and more recently for He recovery from divers breathing
mixtures used in offshore oil production work. It is also clear that the
removal of from, for example, fuel gases, to upgrade their calorific
value, can be carried out easily given the very high permeation rate of
There is, as in PSA, a fundamental compromise which has to be made in
membrane separation: to achieve high purities the membrane must be
absolutely pinhole free and this means that it must be thick. However, to
maximise throughput of gas permeate the membrane must be as thin as
possible. Only recently have selective membranes become available that
satisfy these requirements to a sufficient degree. It remains, however, a
difficult technology; a typical gas-cylinder size separator has up to 1 million
fibre strands, any of which could leak, reducing the assembly performance.
A typical membrane separation unit (Figure 2.2.1) comprises bundles of
hollow membrane tubes or fibres, which can be from 40 to 0.2 mm in
diameter, none too big to pass through the eye of a small sewing needle,
mounted in end blocks inside a cylinder. The fibre manufacture is largely
shrouded in commercial secrecy but available fibres, such as those from
DOW, are based on well-known principles of the asymmetric membrane.
Dealing with pinholes in the selective membrane is the subject of some
secret commercial activity but, to judge from recent claims, has been largely
overcome.
A membrane unit in general works as follows. High pressure raw gas is
continuously admitted to the outside of the fibres at one end of the cylinder
whilst at the other end the depleted waste stream is exhausted from the
fibres. The product is drawn off from the fibre mount i ng block. The hollow
fibres consist of highly permeable non-selective polymer fibres coated with a
thin coating of selective membrane. Typical inlet pressures vary from a few
bars to a few tens of bars.
Figure 2.21 Typical membrane gas separation unit.
In the case of nitrogen generation from air, nitrogen is less permeating
than oxygen. Hence oxygen is the product in the above description and
nitrogen is the waste stream. Clearly, by selection of flow rates, and
perhaps by the addition of purge streams (for greater nitrogen purity, a
purge stream of air down the fibres carrying oxygen permeate away helps),
the process can be optimised. The use of counter-current can also offer some
improvements in performance.
A single membrane separator is often followed by a further membrane
separator in series, to improve purity or decrease product lost, as current
selection factors for single stages are insufficient in many applications. A
cascade of membranes can be used, as in uranium separation, if necessary.
Standard membrane separators are available for nitrogen, with a membrane
selectivity of 6:1 or better. As with PSA, recovery, percentage of the feed
air output as gas, depends on the purity of output required. For example,
with membrane selectivity of 5, and 99% purity, a recovery of 30% is
possible, or with 90% N
2
, recovery of 70%. For a selectivity of 10, which
may be available in the future, recovery would rise to 50% on 99%
For low purities, membranes are showing great promise with comparable
recovery to PSA. Prices for membrane units are still high but are expected to
fall with time. With lower cost membranes in the future, membrane systems
will become more popular, particularly in small applications where their
simplicity is particularly appealing. Isalski (1989) suggests various schemes
for combining membranes with cryogenic separation, using the membrane
to carry out a coarse bulk separation on the feed gas, followed by a cryogenic
unit to give high purity output gases. These more complex schemes will
depend on membrane units reducing in cost considerably, as they are only
make sense for large-scale systems. The Fluor company promotes a fuel gas
+ ethane separation system, using membranes to reject along with
GAS TECHNOLOGY 117
cryogenic distillation, however, so at least some schemes are economic even
today.
Gas separation by centrifuge. Membranes are unlikely to be used for
separation of uranium isotopes in the future as more efficient methods based
on centrifugal separation are available. is put into a set of devices
looking remarkably like spin-dryers except that they are several metres
across and revolve at speeds which would cause ordinary materials to fly
apart. At the many hundreds of g achieved in these machines stage
separation factors several times larger than the membrane method are
available, with the additional advantage of lower losses per stage. Practical
gas centrifuges employ an axial counter-current flow, achieving the equiv-
alent of many stages in each rotating cylinder; in effect the system can be
considered to be a set of disc-shaped elements connected rather like a
distillation column. The feed gas is sent to an appropriate place in the middle
of the cylinder and streams enriched in light and heavy components
extracted from the periphery and the axis of the cylinder at opposite ends.
There seems little chance that a centrifuge method, conceptually simple as
it is, would have anything to offer in, for example, air separation, because
the separation factors with inexpensive centrifuges are too close to unity to
be useful. A simple analysis shows that the separation factor is proportional
to the difference between molecular weights, so separating and which
differ by four units, is barely easier than and differing by three.
Stable isotope separation of gases may be a more practical application for
the technology.
There have been proposals, however, for carrying out distillation in a
column inside a spinning centrifuge. It is easy to show that such a column
could be made a very small fraction of the size of a conventional gravity
column. US military work pursued a project to make such a compact
centrifugal column and this was more or less technically successful.
However, the complexity of feeding liquid cryogens in and out of a spinning
column, and controlling the process, makes it unattractive except for
specialised military applications.
2.2 Making other gases
In many cases, the manufacture of non-air gases is very similar to that of air
gases, except that separation is from an industrial waste gas stream rather
than the atmosphere. Cryogenic distillation of waste streams is rare,
however, because waste streams rarely consist of low-boiling gases like
nitrogen and oxygen. However, PSA and membrane separation methods
are common. In mixtures of hydrogen with other gases, for example, both
PSA and membrane methods are particularly simple and efficient, because
the is so far from the other components, such as hydrocarbons and
in its properties.
2.2.1 Gases from chemicals
Many schoolchildren finish their studies in chemistry under the impression
that gases are mostly made by chemical means. In fact, chemical generation
is not a standard means of production for most gases in industry. It is
nevertheless important.
Chemical generation of oxygen. Chemical reaction-based schemes for
separating oxygen from the air are still occasionally considered today for
special purposes, for example in purifier use. As well as the barium process,
there are possible cycles using high temperature to release oxygen absorbed
at lower temperatures based on nitrate/nitrite changes and manganese
oxide/dioxide changes. MOLTOX is an example of this kind of process,
using alkali nitrate and nitrite transitions at 500 to 650C (Dunbobbin,
1987), although it is still under development. MOLTOX might be very
efficient if integrated with a process (a steelworks, perhaps) which required
operation at elevated temperature.
Less energetic reactions with oxygen have been sought in order to lower
the regeneration temperature in a process. Complexes involving transition
metals (of which, of course, the haemoglobin used in nature is one) are all
possible, including those based on copper (cuprammonium), iron and
manganese. Organo-metal complexes are possible, as are compounds based
on precious metals. Any compounds which form a very weak bond with
oxygen are candidates for a reversible chemical separation process,
although the strong oxygenoxygen bond must be broken and remade,
causing problems for many processes.
Oxygen is occasionally supplied for emergency breathing use via chemical
generation, for example from chemicals such as certain perchlorates or
peroxides which react with water to release oxygen, although the purity and
smell of chemically released oxygen is difficult to control and the gas may
need cooling from some mixtures before it can be breathed. In confined
spaces the release of carbon dioxide, as well as the lack of oxygen, is a danger
to health. Sodium peroxide used to be supplied for use in submarines, for
example, as oxygen candles, as it has the additional advantage of absorbing
carbon dioxide as well as releasing oxygen.
Nitrogen from inert gas generators. The term inert gas generator is usually
taken to mean a system for burning a clean fuel in a precisely metered supply
of air, followed, if necessary, by condensing of water vapour. The inert gas
produced by such an arrangement will vary in its carbon dioxide content with
GAS TECHNOLOGY 119
the carbon in the fuel and will contain impurities reflecting those in the fuel.
However, with good control of the burner, the composition should be 90%
nitrogen with 10% carbon dioxide. The production process is relatively
efficient because around nine volumes of inert gas are produced for every
volume of fuel gas.
Inert gas generators are most commonly used with careful stoichiometric
control of the oxygen supply so that the oxygen content is minimised, whilst
ensuring that little fuel remains in the gas stream. The use of very clean fuels,
methane is most common, improves the quality of inert gas generator output
over that obtained by combusting heavy oils of variable composition and
high sulphur content. Residual fuel, nitrogen oxides, carbon dioxide and
sulphur compounds will all, however, still be present in trace quantities and
may cause process problems or corrode the pipework.
Inert gas generators are only nowadays used in processes with minimal
requirements on gas quality and a steady demand pattern, and then only
when capital and running requirements must be minimised and the size of
gas stream needed is not too large. In the absence of many favouring factors,
the use of cryogenic nitrogen, either on-site or in bulk liquid tanks, is
preferred. In applications requiring a reducing gas, an inert gas generator
can be run with excess fuel. This system is often dubbed an endothermic
generator and is typically used in heat treatment of metals (section 3.2.2).
2.2.2 Carbon dioxide
Smokestack extraction. In some ways the most obvious source of is
exhaust stack gas. Extraction from exhaust stack gases in which, after
is the biggest component is certainly superficially attractive. However,
there are quite considerable problems. The stack gases are generally hot,
dusty, humid, oxidising and highly acidic. Small amounts of sulphur in
feedstocks for most burners result in sulphur dioxide and trioxide, and
sulphuric acid. Some processes produce particularly clean exhaust streams,
some refinery combustion products, for example, and these are generally
preferred over extraction from general furnace and boiler exhaust streams.
Increasingly today, in large stacks, many of these objectionable impurities
are filtered or scrubbed out as an anti-pollution measure. With a cleaner gas
stream extraction is much more economical. However, the way in
which scrubbing is done affects the content of the stack gas.
After prewashing, compression and water separation, a waste stream is
turned into pure via a sequence of processes which varies according to
source and use purity needs. Charcoal beds are commonly used to absorb
remaining contaminants followed, if necessary, by cooling to liquid and
distillation. The condensation to liquid eliminates the large amounts of
permanent gas contamination (nitrogen, for example) and crude distillation
can remove residual unburnt fuel components. The resultant liquid carbon
dioxide is stored under 20 bar or so pressure and at a reduced temperature
(20C) in insulated tanks.
Ammonia plant extraction. Ammonia plants are an important source of
relatively pure carbon dioxide for the industrial gases business. Ammonia
plants need vast quantities of nitrogen and hydrogen in the stoichiometric
ratio of 1 part nitrogen to 3 parts hydrogen. Such large volumes of hydrogen
are a problem. The only economic method is generation via steam/methane
reaction unless a nearby petroleum refinery has hydrogen supplies available.
The reaction produces a mixture of CO and from which fairly
pure must be extracted for the ammonia synthesis. The reaction can be
biased towards and away from CO production by high steam-methane
ratios. Alternatively, the nitrogen required can be generated by partial
oxidation of the methane with air, which can be arranged to leave a stream
which contains nitrogen and hydrogen in 1:3 ratio with mainly
by-product.
The CO and rejection which has to be done provides a natural
by-product stream which, with little further processing, can be liquefied and
stored. The extraction is most often performed by an acid gas wet
scrubber using simple aqueous ethanolamines or specially designed mol-
ecules such as sulpholane. (In gas streams with hydrogen sulphide,
methyldiethanolamine and other tertiary amines are used now as selective
absorbers for which leave the in the stream; this can be convenient,
not least because solutions are highly corrosive to steel plant equip-
ment, forming iron sulphide.) The feed gas is passed up a tower with
counter-current absorber where the is stripped into the liquid. The
liquid is then heated, typically with steam, to remove the and the liquid
recycled. This simple description ignores, of course, difficult details such as
the effect of other sulphur compounds, corrosion, foaming, and degradation
of the scrubbing liquid.
2.2.3 Acetylene, ethylene and unsaturated hydrocarbons
Acetylene. Acetylene is still largely made by the addition of water to
calcium carbide. (There are alternative methods from petrochemical
sources.) The carbide is produced in electrical arc furnaces where calcium
oxide (lime) and coke or anthracite (carbon) are heated to 1850C by arcs
between three carbon electrodes connected to a powerful (multiple tens of
megawatts) three-phase source. The furnace is a metal shell lined with
firebrick and, at the bottom, carbon to resist the red-hot molten carbide.
The furnace can be run almost continuously, with further additions of lime
and coke, while molten carbide can be run off into moulds. The resultant
GAS TECHNOLOGY 121
impure calcium carbide is broken up when cool into the once-familiar
carbide chips. Carbide furnaces emit substantial amounts of noxious fumes,
smoke and noise, and require low-cost electric power, typically hydroelec-
tric. Today, more effort is made to recycle some of the carbon monoxide
evolved when the reaction occurs
The carbide/water reaction can be carried out with a large excess of water,
in which case the reaction occurs smoothly and safely at low temperature.
The by-product, however, is a large volume of unpleasant lime sludge. Some
plants mix a roughly stoichiometric amount of water with the carbide and
apply cooling and agitation to avoid overheating; the pay-off being that only
a small volume of fairly usable impure lime is produced which can be used
once again in the carbide furnace, for example. Various impurities,
including such dangerous gases as arsine, phosphine, silane and hydrogen
sulphide are also produced by the carbide/water reaction, deriving from
impurities in the original materials to the carbide furnace, and must be
absorbed, typically by an acid wash tower to absorb hydrides and basic
impurities, and then an alkali tower to absorb acid impurities such as
hydrogen sulphide. Finally, the acetylene is ready to be pumped into special
cylinders.
The acetylene process has been associated with pollution in a number of
forms and even today is still a potential source of environmental problems.
First, the carbide furnaces are a source of fume and smoke, which are
difficult to suppress. Add to this that carbide furnaces are usually placed
near to cheap hydroelectric power sources, which tend to be in areas of
natural beauty. Finally, the typical by-product of acetylene production from
carbide is a lime sludge, which is difficult to dispose of because it has a high
water content (and thus low lime content) and is often foul smelling as a
result of impurities in the feedstock.
Alternative to the carbide route to acetylene are processes that aim to
convert hydrocarbon feedstock. There is an electric arc process producing
acetylene and hydrogen by heating methane to over 1500C in a multi-
megawatt electric arc for a few milliseconds followed by quenching. A more
typical alternative process adds preheated (600C) methane and oxygen and
burns them, following the flame by a water quench system. With the correct
very rich flame, the reaction products are 50% 25% CO, unreacted and
fully combusted products, and less than 10% This can be washed out
by absorbing it in dimethylformamide in a counter-current scrubbing tower,
leaving the unwanted gases to be used as intermediates or burnt as fuel gas.
The acetylene is flashed off from the dimethylformamide by steam heat.
Absorbers for acetylene in cylinders. Unfortunately, acetylene is unstable
under pressure of more than a few tens of bar, or in liquid or solid form, and
can detonate explosively and disastrously
11
(section 3.2.3). Acetylene
dissolved in an organic solvent under pressure is stable, however. Typically
the gas is dissolved in acetone, which is in turn held in a porous spongy solid.
Organic solvents, such as acetone and tetrahydrofuran, are well known for
their ability to contain large volumes of dissolved gas. The large volume of
acetylene, at 18 barg, which can be dissolved in acetone was a key
factor in allowing the convenient and widespread use of acetylene in welding
and cutting.
12
The stability of the solution is enhanced if the solution is absorbed onto a
microporous spongy substance such as kapok plant fibres, asbestos mineral
fibres, a porous plaster mix or a silica such as kieselguhr. There is a further
complication in that, in order that the acetylene will bubble out of solution
steadily during gas usage, nucleation centres may need to be provided for
bubbles to form. These might be small pieces of charcoal. The porous mass
should also have larger passages in its structure to allow the flow of gas from
remote parts of the cylinder to the valve. A medium with all the ideal
properties is not available so a compromise choice has to be made, perhaps
explaining why there are a number of different alternative materials in
service.
Nevertheless, the search goes on for safer or more efficient dissolved gas
packaging for acetylene (and potentially for other gases). Acetone has a low
boiling point (56C) and a very low flash point, making it a dangerously
inflammable substance to handle. Dimethylformamide has been tried as an
acetylene absorber and, although much more expensive, has a good
solubility for acetylene and is safer. Its high flash point and high boiling point
(150C) make it fairly safe to handle, and its low vapour pressure means that
less solvent is lost by carryover of vapour with the acetylene.
Solid adsorbers are a possibility for acetylene storage although none has
yet achieved any commercial success. Clathrates, specially synthesised
cage-shaped molecules such as the silicate-based molecular sieve materials
or the organic-based hydroquinone compounds, have often been proposed
11
Friable buildings are commonly constructed around acetylene plant. The idea is that in the
very unl i kel y event of an explosion the blast will be absorbed by the disintegration of the
building. Furthermore, often the roof of such a building is designed to offer little resistance to
blast, which thus tends to be dissipated upwards, whilst heavy walls deter lateral blast effects.
12
Acetylene used to be readily available without such specialised gas cylinders; instead calcium
carbide chips were used. This most interesting, if dangerous, compound was sold for putting in
acetylene generators. These consisted of a water t ank with spring-loaded diaphragm
underneath a hopper on the top for the calcium carbide chips. When the outlet was opened and
the pressure fell, the spring-loaded diaphragm relaxed, the hopper allowing carbide chips to fall
into the water. If the pressure increased, the diaphragm closed the hopper once again in a crude
regulation of pressure. A metal version of the laboratory Kipps apparatus would not have
worked too well; the carbide in the middle chamber would have heated up, generating steam,
and pressure control would have been even worse.
GAS TECHNOLOGY 123
as absorbers. However, the cost of these materials relative to their technical
advantage has always so far prevented their adoption. More recently, the
use of the carbon pentagon/hexagon spheroidal carbon skeletons named
after Buckminster-Fuller (fullerenes) has been proposed. Fullerenes,
although only recently recognised, are in fact fairly easily produced, albeit
inefficiently, by dissolving soot from carefully controlled flames in benzene.
The resultant red solution can be evaporated to give a mixture of fullerenes
of around a hundred or so carbon atoms or less. These have some
remarkable properties (for example, superconductivity when suitably
doped with metal ions) and are also remarkable absorbers. Although
currently expensive, fullerenes are composed of carbon and so could
become cheap if efficient production methods become available.
Ethylene. Ethylene is the cheapest unsaturated hydrocarbon. It is
extracted as a by-product from petroleum processing streams, particularly
cracking plants, and is also made by cracking of ethane.
Ethylene is used on such a huge scale that massive plants have been built
to create ethylene from almost any hydrocarbon feedstock. A simple ethane
cracker might simply pass the gas through a counter-current heat exchanger
and then rapidly through a 900C furnace for a fraction of a second. The
output gas stream is then passed back up the heat exchanger to heat more
inlet gas. The exhaust gas will contain ethylene and also methane, acetylene
and other products. It is compressed, when heavier fractions condense, and
scrubbed to remove and other components which might freeze. A
cryogenic separation process at below ethylene boiling point is then used
and non-ethylene components recycled.
Ethylene is occasionally subject to runaway polymerisation in the gas
cylinder, evolving large quantities of heat, or in processes at high pressure; it
is, after all, the monomer of that most common of plastics, polyethylene
(PE), which is formed by applying high pressure. Dry ethylene in a
dessicator in the presence of acid as a catalyst is particularly subject to this
problem. In large quantities, ethylene is delivered in liquid form at low
pressure. However, with precautions, cylinder ethylene is not regarded as a
particular hazard.
Methylacetylene and propylene. There are a number of substituted acety-
lenes and more complex unsaturated gaseous hydrocarbons which have
from time to time found applications. They have lower performance flames
than acetylene but can typically be stored in liquid form and are therefore
more conveniently handled. Petrochemical sources are normal.
Propylene is, like ethylene, an important intermediate in the petrochemi-
cal industry. It is sourced as a by-product from ethylene manufacture. At
high pressures, dry propylene can, like ethylene, easily be subject to
polymerisation. With polypropylene explosive polymerisation can take
place with acidic catalysts but, subject to precautions, it is perfectly safe
under its own vapour pressure at room temperature. It is therefore normally
delivered in liquid form at 10 barg.
2.2.4 Fuel gases: methane, propane and butane
Methane is familiar as domestic piped gas in the more urban parts of the
world. Propane is most familiar in liquid form in low-pressure cylinders for
domestic use, while butane is used in mixtures with propane as the
propellant for spray cans and as the fuel for small domestic burners such as
cigarette lighters.
These gases are all extracted from natural gas or petrochemical streams.
These streams contain carbon dioxide and hydrogen sulphide, which are
typically removed by scrubbing with absorber liquids. This done, the
streams normally consist of a mixture of light hydrocarbons, which require
separation by cryogenic or high-pressure distillation. Ethane is normally
removed, as are ethylene and other unsaturates, as they are more valuable
as chemical intermediates than as fuels. The lightest fraction is mainly
methane and is either piped direct or liquefied for transport by ship to
consumers. The fraction heavier than ethane can be separated into fairly
pure propane, isobutane and butane (with a small admixture of ethane and
pentanes) or can be sold as mixtures. The mixtures are sold as having boiling
points in certain ranges or, which amounts to much the same thing, as having
vapour pressure between certain limits at room temperature. A stenchant
such as a powerful smelling mercaptan or other vapour based on sulphur is
usually added to commercial fuel gases as an aid to leak detection. For
applications requiring complete freedom from impurities such as sulphur
(many chemical processes) or low odour (aerosol can propellant) the gases
are sometimes available without stenchant or else an regenerable adsorber
cartridge purifier can be used to remove the mercaptan compounds.
Methane has the unusual distinction of forming stable solids with water at
comparatively low pressures and comparatively high temperatures (Melvin,
1988). Hydrates such as are formed at 70 bar and 10C, for
example. Methane hydrates cause blockage in high pressure pipelines in
winter and pipelines are often cleaned out with methanol (which decom-
poses hydrates). It is thought that huge reserves of methane gas, perhaps as
much as in current gaseous reserves, are stored in the earth as solid hydrates
but it is very difficult to extract the gas from them.
The fuel gases are nearly always simply burned in air as a source of heat
and are, as a result, largely interchangeable with each other, with other
fuels, from petroleum to coal, and sometimes interchangeable with
electrical or other heat sources. Because of this interchangeability, and
because they are often used in domestic applications, these gases are not
industrial gases in the same sense as the gases such as nitrogen, oxygen or
GAS TECHNOLOGY 125
acetylene. It is for this reason that they are dealt with only cursorily in this
book. There are some significant non-fuel uses for fuel gases, in heat
treatment of metals in furnaces, for example (section 3.2.2). Propane,
isobutane and butane are used in propellant mixtures, mixed with liquid
formulations such as hair spray or paint, in aerosol cans (section 3.8.12).
2.2.5 Hydrogen
Coke oven gas and reformer gas. Hydrogen is a by-product of coke oven
operation. Coke ovens are still widely operated to produce high-quality
carbon for the reduction of iron ore to iron in blast furnaces, although
changes in blast furnace practice, such as the use of large quantities of oil
injection, have reduced coke oven use.
Coke oven gas consists of steam, ammonia and smaller
amounts of many other gases and vapours. After crude washing to remove
the more reactive contaminants PSA units are used to extract the hydrogen
in the stream.
The reaction of methane from natural gas with water, producing the
mixture known as reformer gas, has been in use ever since methane
became readily available. The cleanness of the feedstock simplifies plant
design and stand-alone plants for hydrogen production are common in large
chemical complexes. The clean feedstock also enables the use of more
advanced separation technology, membrane separation, to separate the H
2
.
Reformer conditions can be adjusted to maximise hydrogen yield from the
reactions
The reforming of other hydrocarbons also yields hydrogen, along with
much more CO. Cracking of hydrocarbons also yields hydrogen and this is
sometimes separated in large petrochemical plants.
Hydrogen from methanol and ammonia. The above methods cannot be
carried out effectively on a small scale and smaller industrial plants generally
carry out cracking of methanol or ammonia.
Methanol can be cracked by suitable catalysts into a mixture of carbon
oxides and hydrogen, again followed by separation on membranes or by
PSA, but the latter processes mean that this is only economical at the larger
end of small scale. Similarly, ammonia can be heated to yield a mixture of
nitrogen and hydrogen (cracker or forming gas). This is generally used
directly in processes where the 30% content is unimportant, such as in
bright annealing of steel. Forming gas contains only a few other impurities;
for processes tolerant of nitrogen and those impurities requiring moderate
volumes of hydrogen, cracking units can be a good solution. Forming gas
generators are simple, needing only a controlled temperature hot chamber,
valving systems and some precautions during start-up. Furthermore,
forming gas can often be formed by drip feed of ammonia into a furnace
which is sufficiently hot to ensure complete cracking, again with precautions
during start-up. Bulk ammonia storage and delivery is hazardous, however,
and the relatively volatile bulk ammonia price affects the cost-effectiveness
of forming gas.
Solvay cell and membrane electrolysis for hydrogen and chlorine. Elec-
trolysis of brine yields hydrogen and chlorine, whilst electrolysis of alkali
yields hydrogen and oxygen.
Where electricity costs are low, for example, near large hydro-electric
schemes, and where chlorine must be generated, it is economic to generate
hydrogen on a large scale by electrolysis. Occasionally smaller alkali
electrolysers are used to supply pure hydrogen which would otherwise be
delivered in manifolded compressed gas cylinders, perhaps because cylinder
H
2
is locally expensive or because a strong security of supply is demanded;
some semiconductor facilities have electrolysis equipment, although liquid
hydrogen is now preferred.
Solvay cells electrolyse brine between a flowing liquid mercury cathode
and a carbon anode. Chlorine is produced at the anode, whilst sodium ions
are discharged at the mercury cathode and form a sodium/mercury alloy or
amalgam. The amalgam is then pumped away to a separate cell where it is
contacted with water and graphite, evolving hydrogen from the graphite
surface (which reduces the H
2
production overpotential) and producing
concentrated (50%) aqueous sodium hydroxide. The resulting hydrogen is
substantially pure apart from water and a trace of mercury vapour. Mercury
contamination must be dealt with by passing over iodised carbon filters,
although much of it is removed by compression of the product. Because of
fears of mercury contamination, Solvay cells are now unusual in new plant,
although with suitable monitoring and controls new construction is possible.
The membrane or diaphragm electrolysis cell employs straightforward
carbon electrodes, with a semipermeable membrane, based typically on
asbestos fibre but sometimes a special polymer, between the anode and
cathode. Membrane cells can be used for brine and alkali electrolysis. Brine
electrolysis in these cells has the problem that the concentration of sodium
hydroxide that can be achieved is rather low, little better than 10% and the
alkali is contaminated with unreacted brine and sodium chlorate (from
chlorine) because of leaking across the membrane. Much power is then
needed to extract water from the deliquescent alkali to bring it up to
commercial (50%) strength. Polymer membranes made from cation ex-
change resins can improve this performance dramatically, producing a purer
alkali with less chlorate and chloride contamination.
GAS TECHNOLOGY 127
For simultaneous gas production a cell using sodium or potassium
hydroxide electrolyte is normal.
13
Nuclear submarines employ electrolysis
cells using KOH to generate breathing oxygen during long underwater
voyages. The output is carefully monitored with multiple gas analysers and
with a mass spectrometer to ensure the safety of the atmosphere for the
sailors. The latest systems use membranes made of advanced polymers,
cation exchange resins, instead of asbestos, which allow only sodium (or
potassium) ions to pass, and offer purer products and a more efficient
system.
A surprisingly large amount of electricity is needed to decompose a small
volume of water: about 4kWHr per kilo of water. The low rate of
production is clear from laboratory electrolysers, where a substantial 10
or so power supply yields a weak stream of bubbles. The power is needed at
an inconveniently low 2 or 3 volts d.c. or so. As a result electrolysers are
usually designed in series connected banks so that their operating voltages
come up to levels at which transformers and rectifiers are efficient. The
current electrodes have current densities of from hundreds to thousands
With suitable containment vessels electrolysers can easily produce gas
under pressure and in fact efficiencies can be higher at high pressure with less
blinding of the electrodes by bubbles; compact KOH systems in submarines
take advantage of this.
Electrolysis hydrogen, although very pure otherwise, normally contains
or (and the oxygen or chlorine will contain ), as well as water
vapour. Downstream processing of the gases aims first to remove or
contamination by passing over a heated precious-metal combustion catalyst
(deoxo), then to remove moisture and then residual contamination.
Oxygen from electrolysers is valuable because it is entirely free of carbon
compounds, argon and other inert impurities that oxygen from air separ-
ation normally contains. Some applications, such as in the nuclear reactor
service, which require small quantities of oxygen free from argon, specify
oxygen supplied by the electrolytic route.
Liquefaction of hydrogen. Hydrogen is normally liquefied only on a large
scale for rocket launching, as liquefaction is an expensive process for this
low-cost commodity. Rocket testing and launching continues to drive large
liquid hydrogen (LH) installations, although necessarily long-haul supply
routes in some parts of the North America justify the liquefaction of
hydrogen for transport purposes. One Canadian plant, for example, with
plentiful low-cost energy but far from uses, incorporates a liquefier and
exports the gas to the USA.
Liquid hydrogen is an exceptionally pure product as it will dissolve only
13
Why school demonstrations use dilute sulphuric acid for demonstrations of electrolysis is
unclear; both on a small and on a large scale, NaOH/KOH seems superior.
ppb level impurities at its 20 K boiling point. As a result many semicon-
ductor manufacturers now specify LH for their hydrogen needs. Most
industrialised countries therefore now require at least modest supplies of
liquid hydrogen to be available. Plants for the liquefaction of hydrogen are,
like many cryogenic plants, subject to scaling rules which strongly favour
large installations. There is therefore today a considerable trade in LH to
efficiently use the output from the few large plants, distributing the LH in
cryogenic tanks mounted in standard ISO shipping containers.
Liquid hydrogen is produced by first further purifying a substantially pure
product to remove air gases and moisture, then cooling to 20 K, when
liquefaction takes place, and fi nal l y converting the majority (75%) ortho
form to the para form. Purification via conventional temperature swing
absorption (TSA) systems (section 2.1.5) can guarantee low levels of
and moisture but further removal of air gases may well be necessary, often
with an LN-cooled activated-carbon bed and followed in some cases by an
LH bed or filter just prior to the liquefier.
The liquefaction of H
2
, using Carnot ideal heat engines, requires
12 200 kJ , corresponding to 434 kJ specific heat at 20 K
(6100 kJ work needed) plus a further 6100 kJ cooling from room
temperature. LH thus costs 15 times more than LN to produce, even with
perfectly efficient machinery. The insulation for all liquid H
2
equipment
needs to be of a very high standard; similar to that used for liquid He, with
vacuum insulation frequently being demanded. Safety considerations must
be taken into account with LH plant; roof vents are employed with detectors
to avoid accumulation of there, along with intrinsically safe electrical
equipment where possible.
Small liquefiers, such as those used in research, work in a simple way,
often using a Joule-Thompson valve to expand 100 bar LN-cooled hydrogen
and recycling most of the back to the compressor. Larger, more efficient
liquefaction schemes are used above about 1 tonne per day, with 50 tonne
per day units a typical size. They involve an expansion turbine rather than
LN input to provide cooling to LN temperatures and an ortho-para
conversion at an intermediate temperature such as 77 K as well as at 20 K.
They may also incorporate an expansion turbine for cooling the below
77 K instead of or in addition to the final JT expansion valve.
Molecular hydrogen exists in two forms with the spin 1/2 nuclei (protons)
of the two H
1
atoms parallel or opposite, giving total spin 1or spin 0. (The
isotope is 99.99% of natural H.) With most molecules, the nuclear spin
energies are undetectably close and no consequences are seen. However,
with , the nuclear spin alignment in the molecular electric field
corresponds to energies of 710 kJ , comparable with specific heat
capacities. At room temperatures the spins are randomly assigned, giving
25% para-H
2
and 75% ortho-H
2
. This is the ratio of the number of
same-energy para (nuclear spin 1) and ortho (nuclear spin 0) states that are
possible. At cryogenic temperatures the stable form is the lower energy
GAS TECHNOLOGY 129
para-H
2
, and ortho-H
2
in freshly made LH slowly converts to the para-form,
evolving the 710 kJ heat as it does so. The latter heat causes boil-off and
loss of (ultimately) nearly 50% of the LH over the several weeks that the
transition takes. For transport in tanks, and for many applications, this rate
of boil-off is unacceptable for economic or safety reasons and 100% para-H
2
is required.
Para- can be produced simply by allowing LH to sit in tanks for a while
and recycling the boil-off to the liquefier. A better method is to speed up the
process to a few seconds by employing a catalytic orthopara converter,
usually transition metal salts or oxides. The catalysts can be used near the
boiling point of LH but are more efficiently used in stages, converting a little
more of the stream to para-form at each temperature stage drop in the
liquefier, and hence removing some of the conversion heat at a higher
temperature than 20 K.
2.2.6 Chlorine and fluorine
Chlorine and alkali from brine electrolysers are major components of the
chemical industry. A typical production site is that of ICI in Bromborough,
UK, where hundreds of giant cells are wired in series (to minimise ohmic
losses in the cables) to one of the largest d.c. power stations in the world.
The chlorine produced is purified by drying and then bubbling through
liquid chlorine; this eliminates small amounts of organochlorine compounds
formed during electrolysis. It can then be stored and delivered in low-
pressure (vapour pressure is 6 bara at room temperature) steel or stainless
steel containers as liquefied gas (Figure 2.22).
Chloride and fluoride gases. Chlorine gas is often used to oxidise hydrogen
to make HCl gas, which is sold in small amounts in anhydrous form in gas
cylinders but is mostly dissolved in water and sold as hydrochloric acid. The
combuster is usually a straightforward enclosed flame design, with water
cooling, although materials problems loom large; hot and slightly moist HCl
gas is very aggressive, even on high chromium stainless steel, so carbon and
expensive lined metals are necessary. Hydrogen and chlorine in 1:1 ratio are
fed in and electrically ignited, and the resultant HCl dried, initially by
cooling well below freezing point, to remove moisture. The HCl made in this
direct combustion is favoured for those applications needing high-purity
product. Other chloride gases can also be made by combustion of elements
in chlorine. One example is Solid lumps of boron are burnt by simply
heating to initiate and then regulating chlorine flow to control the reaction
and condensing the (which boils at 12C) produced.
Although known in compounds since Scheeles work in 1771, elemental
fluorine was not produced until 1886 because it is so strongly bound in its
compounds. The element is so reactive and dangerous to handle that it is
nowadays rarely used; one of the few applications encountered in the gas
Figure 2.22 Industrial brine electrolysis cells for chlorine (courtesy of ICI).
industry is the fluorine passivation of stainless steel pipework and vessels.
However, compounds of fluorine such as HF, and perfluorinated gases
such as perfluoropropane are now relatively well-used cylinder gases, and
the CFCs, HCFCs and HFCs used in refrigeration are still gases of major
significance.
Fluorine can be prepared by electrolysis of potassium hydrogen fluoride
solutions in liquid HF. The preparation of HF by dissolving fluorspar in acid
is more important industrially, however, as elemental fluorine is rarely
required. Anhydrous HF made in this way is then purified and liquefied by
refrigeration. Many fluoro gases can be made by using HF in a substitution
reaction with a chloride, using catalysts such as antimony pentachloride
2.2.7 Ozone generators
Cryogenic liquid ozone ( ) is simply a laboratory curiosity. A deep blue in
colour, it is dangerously unstable. Likewise, ozone mixtures in air or oxygen
are unstable, reverting over a day or so back to oxygen. Ozone can be
GAS TECHNOLOGY 131
dissolved in water, it is more soluble than O
2
, but only in relatively dilute
solutions and it is even then typically slightly unstable. Ozone can be stored
stably in solution in chlorofluorocarbons. However, although this is
convenient for laboratory experiments, it is too expensive for any commer-
cial application and in any case CFCs are now regarded as too serious an
environmental problem to be entertained seriously for any new industrial
process. Perhaps studies on the solubility of ozone in more environmentally
friendly HFCs or perfluorinated compounds are indicated.
Because of these storage problems, ozone is therefore prepared commer-
cially by on-site generators. These vary in size depending on application
from less than 1 kg per day to many tonnes per day. Most generators rely on
electrical corona discharge in air or oxygen at pressures from 1 to 2 bara. The
corona discharge is a current-limited high-voltage discharge, typically using
1020 kV a.c. The electrodes are straightforward but must be resistant to
oxidation and need liquid cooling in all but the smallest generators. Another
typical feature of the ozone generators is the requirement that one of the
electrodes be covered with insulator, typically glass. This largely avoids
arcing and consequent damage, although electrodes still fail occasionally.
The alternating voltage is sometimes a straightforward 50 or 60 Hz mains,
although there are some advantages to using higher (audio) frequencies,
which are therefore used by many units. Figure 2.23 shows the principle used
by a corona discharge ozone generator unit.
The discharge forms oxygen atoms by bombardment of the oxygen
molecules by electrons in the discharge. The principal reaction of the oxygen
atoms is to add to an oxygen molecule forming An undesired side
reaction is the formation of oxygen ions, which hinder the electron
bombardment process. Many units use a specially shaped waveform with
millisecond pauses at zero voltage to allow oxygen ion neutralisation,
minimising this side reaction. Once made the ozone will decompose at a rate
which increases with concentration and with temperature. Approximately,
the half-life, behaves as
where = a constant. A typical ozone half-life would be an hour at 100C
and a day at 20C, although ozone dissolved in water, or ozone in the
presence of catalytic substances, may decompose much more rapidly. The
output ozone concentration is limited; with pure oxygen supply, 8% ozone is
possible (highly concentrated ozone can detonate, so is in any case
undesirable) while with air supply, 12% is normal. The feed gas, whether
air or oxygen, must be dry to avoid the formation of corrosive nitric acid.
Electrical efficiencies depend on size of unit and gas supply, with oxygen
units typically twice as efficient as air units. The highest efficiencies are only
around 5 kWHr , which is low compared to the theoretical value of 0.84
kWHr . This inefficiency manifests itself as heating of the gas input,
which is undesirable as the discharge is less efficient at higher temperatures
Figure 2.23 Corona discharge ozone generator unit.
and the ozone formed decays more rapidly. Where an ozone generator is
being fed from a liquid oxygen supply, it is open to the designer to only
partially heat the oxygen gas whose sensible heat, in being warmed from
100 K or so, can therefore be employed to cool the process.
14
2.2.8 Helium
Helium is usually found in natural gases but is usually only economically
extracted in a few parts of the world; for example, in the USA and Poland
where the concentration is particularly high.
Separation of helium from natural gas could in principle be done with a
PSA process, as used for separation of hydrogen from by-product gas
streams. However, in practice, the plants employed hitherto have employed
variations on a cryogenic route. The precise design chosen depends on the
nitrogen and other contents of the natural gas stream.
A typical natural gas contains methane, along with a few per cent of other
hydrocarbons, and up to a few per cent He, up to 50% nitrogen, and
moisture, and The prime aim of the cryogenic process is often to
remove the nitrogen, heavier hydrocarbons and acid gases to make the
natural gas saleable as a fuel gas. In cases where the removed waste stream
contains more than 1% or so of He, it makes sense to extract He as well but
this is the case in only a few places.
The raw methane is washed with an amine solution to remove and
followed by drying over a molecular sieve. It is then admitted, often
under natural pressure but sometimes compressed, to the nitrogen separ-
ator. This cools it and removes the heavier hydrocarbons as liquid, then
14
There is the question of whether ozone can be formed in a discharge in the liquid oxygen
directly and whether this would show improvements in efficiency due to the cooler
envi ronment .
GAS TECHNOLOGY 133
expands it though a JT valve, yielding a mixture which can be distilled to
give methane and an He-containing purge gas. A nitrogen-cycle refrigerator
using a turbo-expander provides cooling for the process.
Whereas hydrogen is only liquefied for special purposes, helium is
normally liquefied because of the intercontinental distances over which it is
transported. There are few sources but many small uses to which it is put. In
addition, many research applications need He in liquid form.
The crude He extracted by natural gas processing must be purified to a
very high standard before liquefaction as absolutely all impurities will freeze
and be potential causes of blockage at liquid helium temperatures.
Hydrogen may be removed in a similar way to Ar processing (section 2.1.5)
by heating in a catalytic combustor after adding excess oxygen. Most
impurities are removed by a stripping process using liquid nitrogen. Use of
LN at low pressure ensures that all N
2
is condensed.
A helium liquefier does not have the ortho-para conversion difficulty that
a hydrogen liquefier has but has the much more serious problem that it must
reach 4.2 K, which is close to absolute zero and is the coldest temperature at
which any industrial process runs. At these temperatures insulation must be
of the very highest order and radiation cold losses can be very serious, so foil
heat shields are necessary where cold parts could see room temperature
parts. The last stage of purification of the helium leaves the gas already
pre-cooled below 77 K at a few bar or tens of bar. Counter-current cooling in
heat exchangers is then used to reach a few degrees above boiling point,
when a Joule-Thompson expansion forms liquid He and gas. Some of the
gas is expanded in two or three expansion turbines; these are small, ultra
high speed devices, with the cool gas cycled back through the heat
exchangers to provide most of the formidable cooling power needed. Figure
2.24 shows the typical scheme used to liquefy He today a cascade of
expanders.
2.2.9 Nitrous oxide
Nitrous oxide is made by thermal decomposition of ammonium
nitrate. The latter requires some handling precautions, as it is potentially a
dangerously explosive compound: often used by terrorists, for example, in
home-made bombs. Inadvertent mixtures with organic or other finely
divided oxidisable materials are extremely dangerous. The nitrate is gently
melted at 170C then pumped into a reactor vessel where is it more strongly
heated to around 250C where it decomposes to steam and nitrous oxide
Subsequently, a condenser separates out the water and cools the gas.
Further treatment is then necessary to remove acidic NO and and basic
impurities such as This is done by successive wash bottles containing
water, alkali and acid. The gas is highly soluble (more than 1 volume per
Figure 2.24 Cascade of expanders to liquefy He.
volume soluble in water), so the wash liquids must be allowed to become
saturated with gas before full output is obtained from the plant. The wet gas
is then dried and stored as a liquid by cooling and compressing either at room
temperature under pressure (50 barg) or refrigerated at lower pressure.
2.2.10 Production of special gases
The myriad of special gases in use in industry means that it is impossible to
describe the manufacture of all of these. A special gas is supplied in small
quantities. For all special gases, therefore, suitably compatible compressed
gas cylinders must be evacuated or purged prior to filling from a bulk source.
For many purposes special purity is needed, above that of available bulk
sources, so small-scale purification is often used. For still other special gases
bulk sources are not available. These gases are then made in small-scale
purpose-built plants. For the more unusual gases it is not economic to run
more than one or two plants and cylinders from these plants are exported
worldwide. Examples of special gases in this category include silane,
nitrogen trifluoride, boron trichloride and phosphine, which are all pro-
duced for the semiconductor industry. Boron trichloride is produced by
burning solid boron in chlorine gas, followed by purification by absorption
and distillation (Figure 2.25).
Some hazardous special gases can be manufactured in-situ, with some
potential advantages in safety, although this is highly exceptional. One US
GAS TECHNOLOGY 135
Figure 2.25 Production of BCl
3
. After burning solid boron in a stream of chlorine gas, the BCl
3
is purified in these absorption columns (courtesy of Epichem Ltd).
company, for example, installed 100 cc gas plants for generating
arsine gas in a mixture with hydrogen. The process is not described
in detail by the firm but presumably involves electrolysis at pure arsenic
electrodes in an aqueous solution, when arsine is mostly evolved along with
hydrogen, followed by removal of moisture.
Making silane. Pure silane for use in the semiconductor industry is often
made by reduction of a chlorosilane with lithium aluminium hydride, the
process being carried out in an organic solvent such as tetrahydrofuran. The
reagents are typically stirred together in a reactor with careful temperature
control, with the whole assembly being located in a deliberately flimsy room,
a friable building, with the operators behind a blast wall. Silane evolved
from the mixture is first cooled, the condensate of chlorosilanes being
recycled, and then cooled further for redistillation to remove solvent and
unreacted or partially reacted chlorosilanes. A typical scale of plant might
make a few tens of tonnes per annum.
15
In the last few years the economics of silane production have been
drastically changed with the arrival of one or two very large plants (hundreds
of tonnes per annum) such as that of DENAL in Japan or Advanced Silicon
Materials, Inc. in the USA. Product from the DENAL and Advanced
Silicon plants is so pure and so cheap that there is now less justification for
small silane plants to produce raw silane. Most gas suppliers now concen-
trate on purifying and certifying silane from these large plants. They use
large tube trailers of high-grade cylinders, other transfilling equipment, gas
analysers and purifiers, so that they can deliver silane of the required grade
from these bulk sources efficiently. The days of small silane plants would
appear to be numbered and silane production will become an oligopoly
similar to the more specialist gases such as arsine, phosphine or nitrogen
trifluoride.
Stable and radioactive isotopes. Gaseous radioactive isotopes are highly
dangerous because they can easily be ingested and cause genetic damage in
humans and animals. At sufficiently low concentrations these hazards are
negligible, however, and gaseous radioisotopes are potentially very useful
tracer gases. The simple apparatus needed for the detection of radioisotopes,
even at exceedingly minute concentrations, commends their use in
applications where immense dilution factors must be dealt with. Similarly,
applications with a need for many inexpensive detectors may find them useful.
There are only a few radioisotopes of the common gases with convenient
properties. Tritium is available at low cost from nuclear reactors, where
neutron absorption by lithium produces tritium and helium. There has been
much research on tritium in the worldwide nuclear fusion development
programme because its reaction with deuterium is the lowest-temperature
nuclear fusion reaction known. Tritium in industry is used as a source of
ionisation in low-pressure lighting tubes, which glow without an external
power source.
Krypton-85 is also used for lamps and also in vacuum tube devices which
require an internal source of ions. Radon-220 is produced by purging a
thorium oxide container with carrier gas and can be used for high speed
tracer work where its exceptionally short half-life is not a problem. Some
radioisotopes used in gas tracer work are shown in Table 2.6.
15
Even more specialist t han silane is the compound disilane which is made only in
kilogramme quantities. LAir Liquide have a patented process for producting disilane from
silane in an electric discharge.
GAS TECHNOLOGY 137
Stable isotopes. Stable isotopes of the common gases are safe to use and
relatively inexpensive. They have a number of uses but tracer applications
probably predominate. By using a mass spectrometric detector, which can
be made very sensitive, many chemical and biochemical studies can be
carried out.
Deuterium extraction is often carried out on large hydrogen-producing
plants with various techniques varying from electrolysis to distillation of LH
being used. Deuterium is the least expensive of the stable isotopes, costing
only a few hundred dollars per kilo (expressed as heavy water), reflecting the
ease with which it is separated. Important in chemical research work,
deuterium is often used as heavy water as a moderator for slowing down
neutrons in nuclear reactors, particularly those used for research.
Deuterated compounds often show very different reaction rates from
normal hydrogenated compounds. A recent study, for example, showed
that vaccines prepared in a heavy water solution were stable for much
longer than vaccines of normal formulation, avoiding the need for refriger-
ation.
Carbon-13 and oxygen-18 are often made by distillation, e.g. distillation
of carbon monoxide, but are much more expensive than deuterium because
their atomic weights differ by only 10% or less from the majority isotopes.
Carbon-13 as carbon dioxide is readily detected in mass spectrometers and is
useful for examining the metabolism of plants, while oxygen-18 labelling can
be used in human studies. Boron-10 gases are used to generate boron-10
beams for ion implantation in the semiconductor industry. Stable isotopes
useful in gas tracer work are given in Table 2.7.
2.2.11 Purifiers
Purification processes may be differentiated from separation processes
because the former are designed to remove only small amounts of unwanted
gas from a mixed gas stream and are not required to provide the removed
components as an output. Absorbers, both liquid and solid, cryogenic
adsorbtion, reactive metal getters and reactive organic getters are possible
principles in purification.
Absorber purifiers rely on washing the product gas with a liquid which
preferentially dissolves the undesired impurities. The most common
washing liquid is, of course, water and it is used in most wet gas scrubbers for
controlling effluent gas (section 2.6.14). Scrubber purifiers generally use a
packed or trayed column to contact the liquid intimately with the gas being
processed. The gas, after passage through the contactor, must be cleaned of
absorber and dried. The absorber liquid, once it has been used for scrubbing
the gas, is usually recycled by passage through another contactor tower, the
stripping unit. In bulk chemical production wet purifier processes are
ubiquitous. Industrial gases manufacture, however, largely avoids the use of
wet purifiers. Generally, the high levels of purity to which the industrial
gases industry works does not allow the use of wet processes. Reference
works such as Kohl and Riesenfeld (1960) provide a wealth of detail of
absorber systems.
Dry purifier principles. Dry purifiers of various kinds are the workhorses
of the gases business. Although designs differ considerably in detail, all have
common features. First, all must contact a solid surface with the gas. The gas
must transfer from the main gas stream to the surface, typically diffusing
through micropores in the reactor bed, then react on the active surface, then
diffuse back through the micropores to the bulk gas stream once again. The
rate at which the purification reaction will take place depends on tempera-
ture, pressure, the area of the active surface and the efficiency of the
diffusion path through the microstructure.
Most purifier materials as well as having a microstructure also have a
macrostructure: the bed material is typically prepared in pellets or granules,
varying in size but most often of the order of millimetres. Sometimes the
GAS TECHNOLOGY 139
pellets are shaped into rings, giving a large amount of free volume for good
flow. Whatever their shape, the pellets or granules are designed to allow the
bulk flow of gas easily and evenly throughout the bed but are small enough to
ensure that the gas can diffuse through to the centre of the granules in a
reasonably short time.
The macrostructure is often complemented by an engineered microstruc-
ture. Instead of trying to prepare a pure form of catalyst in microporous
form, the catalyst is coated onto the pores of a naturally microporous
material. An example would be the use of platinum to coat an inert porous
ceramic such as alumina. The bed materials will often require heating, either
for reaction or for regeneration, and it is helpful to choose support materials
with relatively high thermal conductivities, such as alumina or magnesia.
The active material is coated onto the support typically by soaking in a
solution of a compound of the active material. The ensemble is then
activated by heating to drive off the solvent and then heated, perhaps in a
reducing gas, to react or calcine the active material.
Adsorption purifiers. Dry adsorbtion purifiers rely on the formidable
capacity of substances such as activated carbon (section 2.1.6) to adsorb
small amounts of more reactive gases and vapours from low boiling point
gases and vapours. Cryogenic purifiers rely on the principle that most
impurities have much higher boiling points than the wanted product. By
cooling the product to dry ice or even LN temperatures, many impurities are
simply condensed out. The effect can be further magnified by the addition of
absorbers like activated carbon, which function at much greater efficiency
than at room temperature.
Reactive purifiers contain active reactive chemicals which combine with
impurities but not with the product gas. Small-scale purifiers have a very
simple design, comprising a single vessel loaded with a solid reactive or
adsorbtive material. No provision is made for reuse of the purifier and it is
either jettisoned or returned to the supplier after use. Some of these purifiers
with straightforward active principles can be repacked by the user with fresh
material. Simple cylinders packed with oven-baked magnesium perchlorate
or 13X zeolite granules, for example, can be used to remove moisture from
air or inert gases.
Larger scale adsorber purifiers all employ adsorber materials which are
regenerated in-situ. The two-bed concept is often used for regenerable
solid reactor bed purifiers: one purifier bed is on stream purifying gas whilst
the other is being regenerated by heat and purge gas and/or a regenerating
gas mixture. The regeneration process is preferably a physical process of
desorbtion as this makes for the lowest running costs. Temperature,
pressure or vacuum swings can be used with a product or inert gas purge and
a pressure swing adbsorber purifier capable of the quickest regeneration
time (section 2.1.6). The purge gas is generally run in the counter-current
direction, so avoiding any possibility of polluting the outlet end of the bed
with the impurity.
The simplest industrial adsorber purifiers are the pressure-swing drier
units marketed for drying compressed air or inert gases using the supply gas
itself as purge and power supply (section 2.1.6). After minutes on-line, the
bed (activated alumina is common) is depressurised to atmospheric
pressure, purged briefly and then put on-line once again. These purifers are
commonly operated by a pneumatic timer unit, thus they conveniently need
only gas inlet and outlet connections and a vent; no power or heating inputs
are required.
Temperature-swing adsorption operation is more typical of most purifiers
and usually involves the use of a hot purge gas, perhaps assisted by heaters.
Heating can take a very long time with a large absorber volume so TSA
systems operate much longer cycle times than PSA purifiers; hours rather
than minutes. In addition to the desorbtion process, it is also necessary to
include time for cooling so that the regenerated adsorber is immediately
effective when switched on line.
The temperature needed for regeneration is always kept as low as
possible. First, the rise in temperature costs energy but, more importantly,
the temperature is limited by the adsorber materials. Many adsorber
materials suffer mechanical breakdown (producing fines or dusting) and
loss of capacity as a function of the number of times they have been
thermally cycled; the number of cycles before deterioration is a steeply
increasing function of temperature.
A three-bed system is sometimes used for large purifiers to increase
utilisation of the bed material. This involves having bed 1 in adsorption,
followed by bed 2 in series performing a polishing or additional purifi-
cation, with the remaining bed 3 being regenerated. After the regeneration
cycle is completed, bed 1 becomes bed 3 (regeneration), bed 2 becomes bed
1 (on-line) and bed 3 becomes Bed 2 (polisher), and so forth.
Catalytic purifiers and deoxos. Catalytic purifiers are now a very common
technology; almost every petrol motor vehicle has one to reduce emission of
CO and . Catalytic purifiers in the gas industry are typified by the argon
purifier units (APUs) used to remove the last vestiges of oxygen from argon
derived from air-separation units. Such catalytic units to remove oxygen are
often dubbed deoxos, although reactive getter units which remove are,
confusingly perhaps, also often called deoxos.
The requirement to remove oxygen from inert gases is very common.
Argon from air distillation, for example, always contains some residual
oxygen because the boiling points of the two gases are so close. Argon is
usually used in an application only because it is completely inert; it must
therefore be at high purity (welding argon is an exception in this respect). A
classic design is a heated precious metal catalyst supported on a support,
GAS TECHNOLOGY 141
along with a hydrogen dosing system. After passing over the catalyst, which
might be 0.5% palladium on pelleted activated alumina, the argon will
contain only water vapour, excess hydrogen and trace nitrogen present in
the original gas stream. The system will actually operate at room tempera-
ture, although heating is usual as it avoids certain poisoning effects.
Although generally very satisfactory, catalytic purifiers are subject to
various mechanisms of poisoning by components in the inlet gas stream.
First, particulates or gases which decompose to solids can coat or block the
pores or coat some of the active surface. Second, overtime, the catalyst may
change in physical form, perhaps crumbling into fine dust or sintering
together into large lumps.
Reactive getter purifiers. The reaction of a reactive metal with oxygen in a
gas stream is an important means of industrial gas purification. For small gas
streams containing small amounts of impurity, the getter material is a finely
divided alloy of reactive metals. The alloy is chosen to be very brittle. After
preparation in a vacuum furnace, the alloy is cooled and comminuted in a
vacuum or gas inerted grinder then sintered just below its melting point to
form pellets or larger cylindrical pills in another vacuum furnace. The
pellets are then loaded into a cartridge which is inert purged and shipped to
the user.
After installation, the user simply heats the unit with a simple heater-tape-
type heater (although larger purifier units have an internal heater) for an
hour or so, which activates the getter material. On exposure to air, the getter
material reacts until it has formed an impermeable oxide layer and is then
stable. The heating process activates the getter by allowing the oxygen to
diffuse out of the surface into the bulk of the getter alloy, thus presenting
active surface for deoxygenation action. The purifier can be run hot or
periodically heated to present fresh reactive surface. The capacity of a getter
material for absorbing an impurity, defined with respect to breakthrough of
an impurity at, say, the ppb level, will vary from gas to gas, depending on the
heat of reaction of the gas with the getter. Oxygen is strongly bound and an
oxygen capacity of 20% is typical, whilst for gases such as CO or hydrogen,
which are weakly and reversibly absorbed, the capacity may be as low as
0.6% or 0.4% by weight.
Simple once-through getter purifiers are available for inert gases (He, Ar
and Ne); an alloy of Ti or Zr is typical. The SAES company of Milan, Italy,
are leaders in this field. They have developed special alloys such as Zr/Fe/V
which give improved performance relative to t i t ani um, and removal to ppb
levels, and good capacity for nitrogen, hydrocarbon and as well as
oxygen. The purification of nitrogen with reactive metals is more difficult
than that of inert gases because the most reactive metals also react with
nitrogen and so are unusable. The SAES company has shown that ppb levels
of purity can also be achieved on using a zirconium/iron-based alloy. The
alloy has a composition of 75/25 Zr/Fe alloy, which although it slowly
adsorbs nitrogen to the extent of a few per cent by weight, it remains active
towards impurities. Running at 350C it will remove and hydrocarbons
to 10 ppb or so, as well as removing moisture and oxygen to undetectable
levels.
Once-through getter purifiers are now very common in semiconductor
manufacturing plants, being used as point-of-use purifiers. A point-of-use
purifier is largely used as an insurance policy: it ensures that gas fed at
nominally high purity from a gas distribution system or down a long pipe
from a cylinder placed outside will still be guaranteed to be of high purity.
Changing of filters, reverse flow and changing of gas cylinders can introduce
small amounts of air and moisture, even where the gas supply itself is of the
best quality. Point-of-use purifiers protect against this, as well as upgrading
the purity of the gas.
Once-through getter purifiers are too expensive for high flow rates and for
the purification of gas streams which start off at a lower level of purity. A
typical larger scale reactive purifier therefore has to be regenerable, for
example the supported copper type occasionally used to remove oxygen
from inert gas streams. The copper, spread out on the surface of the inert
porous mineral support, is first reduced fully to copper by a heated
hydrogen/nitrogen mixture. Then oxygen in any gas stream, up to a fraction
of 1%, passing through the purifier will be absorbed in the formation of
copper oxide. After partial exhaustion of the bed over a few hours or days, it
can be regenerated in a few hours by heating with hydrogen. A similar
purifier system that might in the future become popular with semiconductor
plant users is based on a supported nickel bed. The nickel-bed purifier,
originally developed by AIRCO but now marketed by the SAES company,
is aimed at further raising the purity of already good purity cryogenic
nitrogen from contaminant levels of a few or tens of ppm to ppb levels.
Membrane diffusion purifiers. Palladium diffusers are an unusual type of
purifier usable only for hydrogen. The diffuser purifier allows impure
hydrogen under pressure on one side of a heated thin palladium/silver
membrane. Hydrogen dissolves and diffuses through the membrane leaving
impurities behind. In effect, the hydrogen reacts to form an interstitial
hydride with the Pd/Ag alloy, diffuses through the heated bulk metal and
then reverses the reaction on the low-pressure side of the membrane,
forming gas once again.
In principle, the membrane method is similar to the air-separation
membranes described above. Just as with the polymer membranes above,
however, there are problems with leaks. This is particularly the case because
the Pd/Ag membrane undergoes some metallurgical changes with high-
pressure hot hydrogen becoming, like many metals treated with brittle
and subject to cracks in operation. Also, as in other membrane processes,
the membrane must be of the largest possible area and needs to be
GAS TECHNOLOGY 143
exceptionally thin to function effectively, and the maximum possible
differential pressure must be placed across it, giving ideal conditions for leaks
through cracks and pinholes.
Channelling problems in purifiers. The path of gas through a packed purifier
unit is never completely uniform. Whether the packing is a support for a solid
or a liquid absorber or getter, if that packing is uneven, then the gas will follow
the path of least resistance: that which has no packing in the way. In this way, a
badly packed purifier may be found to be useless at all but the lowest flow. Just
as in distillation column design, fabricated mechanical redistributor plates or
additional random structure packing may be useful here. However, an
elegant solution is simply to ensure a very even packing of the column with a
support material which is not too wide in its particle size distribution. (Fines
often tend to build up on larger support material and block it up in an uneven
way.) Channelling effects tend to get worse with size, giving a potential for
difficulties when scale-up of purifier vessels is attempted.
Overheating of purifiers. Another problem that gets worse with size in
purifiers is overheating of the bed inside the reactor. It is often forgotten that
purification absorbtion reactions are usually exothermic, as are regener-
ation reactions.
Models of heat lost from and created in a purifier can be very complex.
However, simple assumptions on the reactor will lead to a parabolic curve
governing the overheat in the middle of cylindrical vessels which is often
similar to that seen in practice
where T is the temperature between the walls, is the temperature at the
wall, is the rate of heat generation per unit volume, S is the thermal
conductivity of the packing, r is the radial position and R is the wall radius.
This means that for small or large S the temperature rise will be small.
However, if is large and S is small (as it is for many mineral support
packings), then the temperature will rise to a maximum along the axis, with
squatter cylindrical or larger reactors considerably worse because of the
factor.
16
The use of higher conductivity packing can be indicated, or the use
16
A new design of large purifier that the author once set up was subject to overheating problems
which only surfaced after an extended period. The unit concerned was of the dry-getter type,
where the getter was regenerated with a gaseous mixture. In normal service, the unit ran for
some hours purifying, then switched for a shorter period to regenerate. All went well until one
day the source gas deteriorated badly and was fed to the purifier rather than being vented. The
purifier worked hard to purify the off-spec gas and then switched over to regenerate. Then the
regenerate gas began to flow and added its exothermic reaction heat to that of the usual heaters.
The temperature rose and continued to rise, at least in the middle of the reactor bed, despite the
thermostat cutting off the heaters. The getter was finally overheated and sintered together. The
purifier was ruined, although the only external sign of this was a small plastic control pneumatic
signal which had melted in the radiant heat from the reactor vessel.
of multiple parallel beds, where temperature rises are thought likely to be
troublesome. However, a preferrable solution in computer-controlled units
may well be modification of the software. By allowing a slow start to
regeneration, perhaps using a diluted regeneration gas, for example,
overheating may be avoided in that part of the cycle, whilst overheating in
operation might be avoided by sensing impurity levels and reducing mass
flow through the unit when levels are high.
2.2.12 Preparation of gas mixtures
With so much effort going into separating gas molecules, it may be a surprise
that a highly profitable, though small, part of the gas industry is involved in
mixing gases together again. In larger quantities, a gas mixture is most often
produced at point-of-use by mixing pure gases with a mass flow controller or
pressure regulator-based rig. Such a rig is not quite as straightforward as
might be imagined, however, for example a turbulent mixing vessel may be
necessary at the injection point of one gas into another. Also, accurate,
calibrated flow controls are needed, which function to give an accurate ratio
of gases at different flow rates, and there may be problems during start-up or
shut-down, with off-spec mixture being briefly produced. Because of these
problems, good mixing rigs are relatively expensive for applications needing
small quantities. Cylinders of compressed gas mixtures are the best solution
where modest supplies are needed, offering convenience, reliability and low
capital cost.
In principle the preparation of cylinder gas mixtures is very straight-
forward. However, some precautions are needed to ensure that the mixture
is safe, is actually mixed (and will not condense and unmix) and is accurate.
Safe gas mixtures. Clearly some mixes are completely impossible; a
request for a mixture of ammonia with hydrogen chloride, for example, will
result in a cylinder containing only a few parts per million of the mixture (at
equilibrium) and a large lump of solid ammonium chloride, as the gases
spontaneously react.
Some mixes, typically those which contain an oxidiser and a fuel within
their flammable l i mi t s, are even more dangerous because they are perfectly
stable unt i l ignited, when a dangerous explosion will ensue. The prediction
of flammability at high pressures is not straightforward. Oxygen mixes, for
example, are often more dangerous at high pressure, with much wider limits
of flammability.
Other mixtures, those containing reactive gases at low levels diluted in
inert gases, will require special cylinder preparation. Over a period of
months or less, components like NO or HCl in a gas may react with the
cylinder walls or contaminants on them and may disappear altogether.
Although al umi ni um cylinders are intrinsically lower in reactivity with most
GAS TECHNOLOGY 145
gases, even Al cylinders need special preparation; drying by heating during
evacuation, for example, prior to filling. For ultimately stable mixtures, alu-
minium cylinders which have been polished followed by addition of a con-
trolled oxide passivating layer (to replace that formed spontaneously in air)
offer one of the best solutions. Anodising can be used to provide oxide passi-
vation of aluminium cylinders. Proprietary (some nickel-based) coatings are
available for improving the performance of steel cylinders. Cylinders can
also be passivated against reaction with fluorine-containing mixtures and
other highly reactive mixtures by a controlled exposure to diluted fluorine.
Using passivated cylinders, gas mixtures such as 50 ppm in nitrogen,
which are proven stable after five years, can now be supplied.
Filling procedures for mixtures. The simplest way to fill an accurate gas
mixture with a mixture of non-condensing gases is to fill by pressure. The
smallest component by volume is added first, then the next smallest constitu-
ent, then the balance gas.
Often greater accuracy is needed, however, and this is where accurate
gravimetric filling is called for. A fundamental difficulty here is the small
relative weight of gas which must be measured. A 100 kg cylinder will con-
tain only around 10 kg of gas at typical molecular weights and only 0.5 kg
hydrogen, even at full cylinder pressure. However, a 100 ppm component
may need to be measured to 1%, which means that 10 mg must be measured
on a 100 kg cylinder, requiring a very accurate balance. A further funda-
mental difficulty here is the effect of the gas connection on the scale reading.
A flexible pig-tail of pipe is necessary, and care must be taken so that the
pig-tail cannot cause inaccuracy. After filling, a more accurate gravimetric
check measurement can be made with the pig-tail disconnected. The gravi-
metric method can be further extended to low-level mixtures by further
gravimetric dilution of a premixture of (say) a gravimetric 1000 ppm mix-
ture; low ppm mixtures can be reached with gravimetric accuracy with a
relatively simple weighscale.
Standards laboratories, such as the National Physical Laboratory (NPL)
in the UK and the National Institute of Standards and Technology (NIST) in
the USA, produce gas mixtures to internationally agreed best procedures
known as standard reference materials (SRMs). Gas mixture manufacturers
may offer traceability of their gas mixtures to the SRMs as an alternative to
cylinders prepared to traceable weight standards via a gravimetric produc-
tion route.
Low-level mixtures of vapour in gas are most conveniently prepared by
injecting the vapour in liquid form with a calibrated syringe into the gas
stream going to the cylinder.
Ensuring homogeneous mixtures. The simplest means of mixing cylinders
is simply to lie the cylinder down (section 1.3.10). However, for speed of
Figure 2.26 Theoretical worst case moisture in a cylinder gas versus cylinder pressure.
operation and to ensure mixing of very heavy and very light gases, rolling is
often specified. This involves placing the filled gas cylinder horizontally on a
pair of electrically driven parallel rollers. After only a few minutes even a
mixture of a very heavy gas with a very light gas, such as with hydrogen,
will be thoroughly mixed. The horizontal attitude in any case ensures
diffusional mixing will take only hours in a normal-sized cylinder and
rotation at 60 rpm or so invariably brings the mixing time down to well under
an hour.
For large quantities of a gas mix, a 5 m long torpedo or tube cylinder
with a capacity of around at 175 bar pressure may be used. This can
be filled, mixed by rotation and then decanted into ten or so standard-sized
cylinders for delivery, with the large cylinder used for mixing never leaving
its horizontal rollers. For mixtures such as those for welding, which do not
need to be very accurate and are used in large quantities, filling and handling
in the vertical position is convenient. For this, it may be an advantage to fit
dip-tubes to the cylinders. A dip-tube tends to give a more effective
circulation in the gas cylinder, particularly if light gas is injected into the
heavy gas. Attempts have been made to devise special designs of dip-tube
which further enhance mixing in the upright gas cylinder, for greater
convenience, such as the patent by Jolley (1990) describing a venturi
attachment to a half-length dip-tube.
Moisture in gas cylinders. A frequent source of complaint about cylinders
of gas mixtures concerns varying levels of moisture as the cylinder is
consumed. It is always worth remembering that in a typical gas cylinder a
small amount of moisture is always present and this moisture will tend to
become concentrated at low pressures as the vapour pressure of the water
becomes a larger fraction of the total gas pressure. The (hopefully rare)
GAS TECHNOLOGY 147
example of a cylinder that has accidentally been partially filled with water
prior to filling with compressed permanent gas is shown in Figure 2.26. This
illustrates the moisture content at the outlet, as a function of cylinder
pressure, assuming the water vapour stays roughly in equilibrium.
2.3 Gas measurement and analysis
Many gases must be measured in quantity and quality for their suitability for
a process. In addition, most articles of commerce have to be measured or
counted so that they can be correctly accounted and charged for. Today,
also, materials which may end up in the environment after use, or as a result
of accidental release, must be monitored before discharge and in the
environment. Gases are no exception to these requirements but they are not
so easy to measure. The fundamentally intractable properties of gases,
invisibility and intangibility, mean that human senses are not useful.
Instrumentation of a more or less sophisticated level is essential to work with
gases in all but the simplest applications.
2.3.1 Gas flow measurement
Simple measurements are often not difficult with gases. Gases can, like most
substances, be weighed in containers although, as noted in section 2.4.7, the
container weight is high, which causes accuracy problems. Pressure
measurements are also fairly straightforward. If access to the gas container is
available, then gas volume can be often be more accurately measured there,
via pressure and/or weight measurements, rather than by instruments in the
pipeline.
For compressed permanent gases, unless accurate scales are available, it is
simpler and reasonably accurate to measure pressure and use the ideal gas
law (with a table of compressibility Z values if available). Volumes of gas
containers are known or can be measured by weighing empty and full of
water. If accurate measurements are made of pressure, the contents of
compressed gas cylinders can be accurately given by pressure, and it is often
forgotten that observation of the fall in cylinder pressure is capable of giving
an accurate average flow rate in many applications. The advent of accurate,
relatively drift-free electronic pressure gauges means that flow rate via
display of electronic rate of change of pressure can be considered; similarly,
flow rate via electronic rate of change of weight is perfectly possible,
although apparently rarely done.
However, access to the gas container is often not available, and in any case
a more direct measurement of instantaneous flow is usually needed. Flow
measurement of gases at an approximate level is not difficult, as the simple
rotameter shows. However, accurate flow measurement is much more
Figure 2.27 Venturi and orifice plate meter systems.
difficult and few techniques are capable of much better than 1% accuracy
unless elaborate precautions are taken, as in laboratory tests.
Venturi and orifice meters. Probably the simplest way to measure a gas flow
is a to include a venturi or an orifice plate (Figure 2.27). The flow is related to
the square root of the pressure drop across the device. The venturi can be
regarded simply as a low-loss orifice and both can be treated by similar
theory.
This arises simply as a consequence of Bernoullis equation (section
1.3.5). First, assume that the venturi or orifice is small compared to the
general pipe diameter. Then the velocity in the throat of the device, will
be considerably larger than and therefore may be neglected
Taking a small pressure drop, with density approximately constant and
Now noting that Q, the volume flow, is given by
Note that the Q value obtained also depends on the absolute pressure of the
venturi/orifice via the density. Note also that Q will vary from gas to gas,
depending on the gas density. The constant is a correction factor for the
particular geometry of venturi and is close to 1 for good venturi devices. is
less than 1, typically 0.65 but varying with Reynolds number, for orifice
plates.
GAS TECHNOLOGY 149
Figure 2.28 Rotameter.
Venturi meters are preferred when accurate measurements are made, or
when a low net pressure drop is required; a good venturi will have very little
energy loss as the pressure lost in the constriction is nearly all replaced in the
diverging section downstream. Orifice plates are used for convenience; they
can be inserted into a flanged joint, for example, quickly and at minimal
cost. They have the disadvantage that they induce a high pressure drop and
are not in general so accurate.
Rotameters. The rotameter or floating ball gauge is very common, despite
the increasing usage of electronic flowmetering. The word rotameter
includes the rota prefix to indicate that it is a meter with a rotary principle.
Accurate rotameters employ a bullet-shaped float with small angled flutes. If
flow is present, the float rotates, showing that the meter is working and not
jammed.
The basic concept of a rotameter is very simple: if gas flows, it forces the
float up the slightly conical inside of the sight tube until the annular gap past
the float sustains a differential pressure equal to the weight of the float
divided by its area.
A simple rotameter can show a roughly linear relation between float
position and volume flow rate (Figure 2.28). To see how this can arise, the
formula above for the venturi/orifice meter is first noted, using the annulus
area between float and tube as the effective orifice
For a float of mass M, radius
or
For a tube with a conical shape, where the position of the
float is x. Hence
Note that the dependency on the gas density means that rotameters must
be calibrated for different gases.
Although simple and popular, rotameters have a number of problems in
practice. The sight tube must be accurately made and has to be glass to avoid
the wear and thermal expansion problems inherent in other transparent
materials. A glass tube with high pressure inside is potentially very
dangerous so a rotameter should generally have a safety tube of clear, strong
plastic, such as polycarbonate or polyester, surrounding it. The glass tube
needs to be sealed into the pipework system with O-rings, with precautions
to avoid leaks. The glass tube is itself not objectionable from a chemical
point of view in nearly all gases. However, the polymer O-ring seals can be
attacked by reactive gases and the possibility of the glass tube leaking a
hazardous gas after breakage is also a problem.
Common designs are limited to 10 bar or so pressure. Some of the
problems can be avoided in designs where the position of the rotameter float
does not have to be sighted optically. One common design has a magnetic
float, which moves an external indicator needle and can operate up to 200 or
300 bar. Another limitation of rotameters is the limited flow range over
which they are reasonably accurate. A typical unit can only read to a few per
cent over a flow range from 20 to 100% of full-scale. Some rotameters are
fitted with two floats of different materials. At the bottom is a high-density
steel float whilst above it is a lighter glass or aluminium float, which extends
the range available by a factor of the square root of the float density (for
identical float dimensions).
Positive displacement meters. If a gas is at a known pressure, then pushing
it (gently, so that the known pressure is not changed) through a cylinder/
piston or similar positive displacement machine will measure its flow; each
revolution of the machine will meter out a known volume of gas.
This is the principle of domestic fuel gas meters and is also widely used in
industry where it is practicable as it is both reasonably low in cost in small
sizes and accurate. Bellows seals are used, to avoid possible wear and gas
leaks via sliding seals, and these meters can be made at surprisingly low cost.
However, this kind of meter has quite a number of problems:

high mechanical wear and low accuracy when used at high flow rates;

large sizes difficult and expensive;
the bellows used are subject to damage with pressure differentials;

normally only usable at low pressures;

normally only usable on gas near to room temperature.
GAS TECHNOLOGY 151
Figure 2.29 Swirl plate turbine flow meter.
Wet versions of these types of meters, for instance those using vanes
sealed with water, are used in laboratories and can be made very accurate.
However, adding water or oil vapour to the gas, which may then need to be
removed, is undesirable in industrial use. Sliding vane meters as used in
liquid metering, where the vanes are lightly spring-loaded so as to present a
minimum resistance to flow, are also used and can be made accurate.
Turbine meters. In larger industrial installations the turbine meter is very
important. A typical turbine meter consists simply of a rotor, often
somewhat like a high-pitch marine propeller in form, rotating on an axis
down the middle of a special pipe section. The rotor may simply operate a
meter display via reduction gears. Alternatively, the rotor may contain a
magnet which gives electrical pulses as it passes a pick-up coil mounted on
the pipe section wall; this arrangement minimises possible mechanical
problems. The bearing used for the rotor is crucial as it must remain
absolutely free-running, even in the prescence of reactive gases and
adventitious impurities, such as water, which may exacerbate the corrosivity
of the gas being monitored. Tungsten carbide or other ceramic ball-bearings
are sometimes used.
The swirl plate turbine is a particularly simple form of the turbine meter
and can be made highly accurate (Figure 2.29). It comprises a fixed stator
which imparts a rotation to the gas. This rotation is then sensed by a thin flat
vane rotating about a central axis. The low inertia of the vane relative to a
standard turbine impellor means that the swirl plate turbine responds more
quickly to changes in flow rate.
This principle has been used in an electronic spirometer for human
breathing measurement. In this, the vane is an ultralightweight plastic type,
metal coated. Its rotation is sensed by a photodiode which is illuminated by
an infrared beam reflected off the vane from a light-emitting diode.
Vortex shedding meters. Vortex meters work by placing an ingeniously
unstreamlined body in a gas stream. As the gas flows, it generates a regular
series (a von Karman street) of eddies; vortices which are carried along
with the stream behind the body. It so happens that, provided the stream
velocity lies between certain limits, the number of eddies is precisely
proportional to the volume of gas that has passed. The eddies are normally
detected with a small piezoelectric microphone or with small anemometer
hot wire sensors. Although technically interesting, they are rarely used in
the gas industry.
Correlation flow meters. In a sense related to the vortex shedding meter
above, in that they rely on turbulence in the flow, correlation meters
measure a gas parameter, such as instantaneous pressure P(t), at two points
x and y, and correlate the signal at the two points, one upstream of the other
by a short distance. The correlation is performed by a small microprocessor,
based on a simple algorithm, such as
choose
calculate correlation sum: (summed over a sequence of
100 or so t values)
choose different values of
find maximum
output (corresponding to maximum
The correlation summmation has a maximum when the offset time, is
such that the two signals have a similar time sequence. essentially gives
the transit time of flow noise from x to y, and, knowing the distance x to y and
the pipeline flow characteristics and cross-section, can be used estimate
volume flow.
Thermal mass flow meters. The thermal mass flowmeter, often abbrevi-
ated to MFM, is now one of the standard forms of flowmeter used in the
gases industry and in some user industries (Figure 2.30). The MFM is often
used as the sensing element in a mass flow controller (MFC). The thermal
mass flowmetering element is followed by a solenoid control valve which is
operated by an electronic feedback circuit such that the flow through the
valve is regulated to be equal to the set flow required. MFCs are now the
standard instrument for the regulation of relatively small gas flows in gas
applications in semiconductor manufacture.
Most MFMs rely on two resistive windings around an insulated small bore
stainless steel tube. Essentially, the windings heat the tube up. In the
absence of flow, the two windings on either side will be at the same
temperature and therefore the same resistance. These windings are
connected into a bridge circuit. Once flow begins, the temperature of the
downstream resistance winding will increase whilst that of the upstream
GAS TECHNOLOGY 153
Figure 2.30 Thermal mass flow meter construction.
winding will decrease, unbalancing the bridge circuit. The unbalanced
voltage across the bridge, if the construction of the sensor is correct, is
approximately linear with the thermal mass of gas flowing. Gases of
different thermal capacity will give different results but this is easily allowed
for.
For flow rates which are large, above about the simple tube
sensor becomes nonlinear and inaccurate and presents too large a pressure
drop. This problem can be handled by arranging a bypass flow path so that a
small but constant fraction of the flow bypasses the sensor tube. The
approach generally taken is to operate both the sensor and the bypass in the
linear flow regime by using very small effective tube diameters (and many
tubes in the bypass circuit).
The thermal MFM has wide applicability: its wholly stainless steel bore is
inert to most gases, its output is independent of the supply and outlet
pressure, its construction is such as to give good long-term stability, and it
can be built to have a fast response time.
Thermistor/hot-wire anemometers. The thermistor or hot-wire anem-
ometer is frequently used to measure gas velocity but it can also be installed
in a pipe section to act as a flowmeter (Figure 2.31). It is also a convenient
instrument with which to probe gas flows in reactor vessels, for example. Its
advantages are that it gives gas flow speed irrespective of orientation of flow
and over a wide dynamic range. By using tiny pieces of tungsten or platinum
wire, and by using miniature bead thermistors (sizes down to about 1 mm
square are readily available), a high speed response can be obtained; as little
as milliseconds. Also, by using an array of these tiny sensors, or by scanning
them mechanically, gas velocity can be probed on a scale of millimetres.
Figure 2.31 Thermistor anemometer principle and circuit.
The most common form of thermal anemometer utilises a feedback
electronic circuit to maintain the sensor element at constant resistance (and
therefore temperature). The parameter measured is the power needed to
maintain this set-point elevated temperature. The elevated temperature can
be anything from a few C above room temperature for a thermistor to
hundreds of C for a hot-wire sensor.
At extremely low flow rates, there is some distortion of the response due
to convection currents from the sensor. However, at even small flow rates
the forced convection response is seen; it is nonlinear but can easily be
calibrated. Various forms of equation can be used for the heat loss from the
wire or thermistor but they all give approximately the same fourth power in
voltage applied result for the flow velocity
rate of heat loss,
where A and B are constants, U is the gas velocity and is the gas density. B
is fairly small in normal operation. Noting that
where V is the voltage applied and R is the (constant) resistance of the sensor
element, we can convert the equation to give the electrical response to flow,
i.e.
Temperature compensation is necessary, particularly for the lower
temperature thermistor-based anemometers. A change in the gas tempera-
ture will change the values of A and B. The selection of wire or thermistor
element is made on criteria of high-temperature coefficient of resistance
GAS TECHNOLOGY 155
(this increases the accuracy of the temperature regulating circuit), corrosion
resistance and resistance value to match the supply voltage chosen.
Ultrasonic flowmeters. The blood flow indicator used in medicine is an
example of a Doppler ultrasound meter. It uses 2 MHz soundwaves reflected
from particles in human blood to give a Doppler shift to the reflected signal,
which gives a measure of the blood velocity. Ultrasonic meters for gases
cannot function in this way because gases contain little or no particulates in
suspension and must rely on the intuitively straightforward time-of-flight
technique or an equivalent phase shift. Meters have been made that will
work in pipework down to 10 mm or so using 100 kHz signals and phase shift
detection. However, the technique is still only applied to gases rarely and
costs are high relative to more traditional flowmetering methods.
Two tranducers are needed, typically mounted in a special pipe section
which allows smooth flow of gas while allowing ultrasonic pulses or
modulated waves to cross the pipe diagonally. The transit time in both
upstream, and downstream, directions is taken together with the
distance apart, D, and the resultant velocity, calculated
is very nearly the average gas velocity, with the effects of gas composition
(speed of sound varies as largely eliminated. There
are quite a few problems with such a system in the signal processing. First of
all, a pulse applied to an ultrasonic transducer produces a rather messy
signal because of resonances in the transducer and its mounting. Secondly,
the received pulse is even more messy due to multiple reflections and
conduction along the pipe walls (the speed of sound in the steel of the pipe is
much higher), and it is difficult to define arrival time precisely. However, the
arrival time changes only slightly with normal gas velocities and it must be
measured precisely. Inexpensive digital signal processing hardware allows
sophisticated processing to alleviate these problems. Gated-phase measure-
ments on the transmitted and received pulses, for example, can be used; the
measurement uses phase-shift on a portion of the received waveform which
is large enough to be clear of small contaminating signals but early enough to
avoid the larger multiple-reflection signals.
The use of ultrasonic devices on larger pipes is now well-established:
larger pipes are technically easier to measure. On larger pipes, too,
measurements over four or more paths across the gas stream can be made by
using three or four transducers. This allows corrections for changes in radial
flow velocity profile. However, with the increasingly lower cost of process-
ing capabilities, it is to be expected that ultrasonic flowmetering could
become more common in small-bore pipe systems. Figure 2.32 shows an
ultrasonic flowmeter intended for low-pressure methane service.
Tracer techniques for gas flowmetering. Some of the most accurate
methods, achieving as good as 0.2% in the laboratory, rely on the use of
Figure 2.32 Ultrasonic flowmeter.
special tracer techniques. Radioisotopes can be used as the levels needed are
low enough to be allowable under health and safety guidelines. Krypton-85
is typical, its emissions being readily detectable with a plastic scintillator/
photomultiplier combination. Other gases can be used as tracers, an unusual
noble gas for example, with a sensitive mass spectrometer detector. Even
ionisation has occasionally been used in tracer-based flow measurement.
Tracer techniques can employ either the steady-state dilution method or
the pulse-injection method. In the first case, the concentration of a known
flow rate of tracer is measured well downstream of injection. The known
flow rate is then divided by the concentration to obtain the pipeline flow
rate. In the latter technique, two detectors, one well downstream of
injection, the other even further downsteam, constantly monitor the tracer
concentration in the pipe. Small pulses of tracer are added and these give
rise to peaks of signal at different times at the two detectors. The distance
between detectors, L, the interval between the half-height of the peaks,
and the cross-sectional area, A, of the pipe yield the volumetric flow rate, V
Both methods rely on achieving thorough mixing across the diameter of
the pipe. Steady-state dilution also has to rely on a known mass flow rate of
tracer, however, whereas pulse-injection depends only on the physical
dimensions of the pipework. Pulse-injection is also more suitable where the
tracer has a very short lifetime, as is the ionisation injection technique.
Measuring gases as liquids. Liquid nitrogen and other cryogens can be
metered in a number of ways but are often not straightforward to measure
because of a tendency to boil and flow in both the liquid and gas phases.
Some of the most common techniques for measuring cryogens are:
GAS TECHNOLOGY 157
buoyancy gauges in small tanks with top access;
differential pressure (DP) gauge on tanks (this is not very accurate on

large flat tanks);
capacitance level gauge on tanks (this is also not very accurate on flat
tanks);
rotary turbine meters (LN can boil inside them and cause them to rotate at
very high speed, albeit briefly);
calibrated venturi meters;
load cells indicating tank weight.

Gravimetric measurement of tanks contents is often used as a gold
standard to calibrate other means of flow and level measurement.
However, load cells are inconvenient and expensive (when accurate ones
are required) and are subject to errors during stormy weather because of
wind and rain. Where there is a nearby weighbridge, tankerloads of liquid
cryogens can usefully be weighed before and after delivery as a means of
calibration. It might be thought that the ideal way to measure cryogenic
liquids is to use a vibrating loop type of Coriolis mass flowmeter. In
principle, these would be ideal, compensating precisely for the presence of
bubbles in the two-phase flow. However, these are expensive and appar-
ently do not fulfil their promise in practice because there are little or none in
service in the industry.
Differential pressure tank gauges. The differential pressure on a liquid
cryogenic tank is in general a very good measure of contents. With an
ordinary liquid tank DP is not particularly convenient. Two single-ended
gauges must be used and the readings subtracted electronically or the
pressure difference between the vapour space and a constantly bubbling
dip-tube must be taken. However, with cryogens a single differential gauge
can be used, with its two inputs connected to the top and bottom of the tank,
because the two lines can be kept clear of any liquid by arranging for them to
be uninsulated when the cryogen boils off leaving gas. Most cryogenic tanks
have a high aspect ratio, length to width, and have a constant cross-section
for over 80% of their length. They are most often placed vertically.
Furthermore, tanks are always run with ullage space in the top dished end.
These facts all conspire to make a DP gauge a useful device for a cryogenic
tank. The variation in density of the liquid in storage does not affect the
measured differential pressure
where is the liquid density, H is the height of liquid in the tank and g is the
acceleration due to gravity. If the lower pressure tapping is arranged to be
above the lower dished end, then the mass of liquid stored in a tank of area A
(lower dished end contents ignored) is
Note that the lines from the tank to the differential pressure gauge must be
full of gas to get an accurate reading. The bottom line should be left
uninsulated and a restriction such as a needle valve or small orifice can be
fitted where the line leaves the tank to guarantee this.
Small cryogenic tanks are often fitted with crude buoyant float gauges. In
these, a sprung cylindrical float rod is arranged to project into a small
transparent plastic dome with graduations on the top of the tank. When full
the float is forced ful l y up against the spring, whilst when empty its weight
falls on the spring and the indicator falls to the bottom of the graduated
dome. These small cryogenic tanks are also frequently fitted with more
accurate capacitance level gauges. These simply measure the capacitance of
a pair of coaxial tubes dipping vertically into the cryogen. The capacitance
increases with the amount of cryogen in the annular gap and is measured by a
simple electronic circuit.
17
Although less accurate than the DP gauge, these
gauges are convenient to fit and inexpensive. They could potentially be
offered with an electronic differential output indicating flow directly but this
is not an option on devices currently available.
It is unfortunate that one class of tanks for which accurate gauging would
be very desirable, the tanks of road tanker trucks, is not so tractable. In
Germany, road tanker trucks are fitted with carefully calibrated venturi
meters. In the UK and elsewhere turbine meters are used but are often less
accurate than a weighbridge.
Gauging high-pressure liquefied compressed gas. Another class of gas
storage where contents indication is difficult is inside compressed gas
cylinders containing liquefied gas. It would be useful to have a level
indication on cylinders containing or propane, for example; the
pressure gauge on such cylinders only indicates a drop when the cylinder is
cool or if it is very nearly empty. Currently, almost all users of these gases
have to rely on weighscales.
Special propane tanks, for example those used in vehicles (or even hot air
balloons), can be equipped with a mechanical float gauge connected to an
indicating needle which is viewed via a small windowed binnacle on the side
or occasionally with remote readout. However, cylinders are at a much
17
An even simpler older device immerses a resistive wire into the cryogen. The wire is heated
slightly by passing a current. The part immersed in cryogen remains cold and less resistive whilst
the part above the level heats and rises in resistance (by a factor of 10 or more if it is heated to
near room temperature) so the higher the resistance, the lower the l i qui d level. The wire is
typically mounted inside a sheath tube (open at top and bottom) which protects it and preserves
its heat to some extent. It is possible that this type of level gauge could reappear in common use
because now the simple wire can be replaced by a length of high superconductor,
which would avoid some of the heat dissipation by the wire, that in the cryogen, and be more
accurate. One such device, at least, is now available commercially.
GAS TECHNOLOGY 159
Figure 2.33 (a) Bourdon and (b) capacitance pressure gauges.
higher pressure (60 barg as opposed to 10 barg) than propane and such a
device penetrating the cylinder wall is unacceptable. High-pressure versions
of the capacitive gauges for use with and other high-pressure liquids do
not seem to be available currently, perhaps because of difficulties in
satisfying the need for ultimate robustness (the gauge head would need to
remain on the cylinder during handling, filling, etc.) and necessary safety
standards.
2.3.2 Pressure measurements
Bourdon gauges. Pressure measurement in industry still commonly uses
the traditional Bourdon gauge where no particular accuracy is needed and
where a gauge pressure (relative to atmospheric) is needed.
18
(Figure 2.33)
A tube which has been curled into a spiral form when filled with high
pressure fluid tends to uncurl. This uncurling movement is linked to a
magnifying mechanism which drives the indicator needle across a dial.
Dead volume (the volume of gas which is stored up in the gauge) is a
problem with Bourdon gauges. Real dead volume is the actual volume of
gas trapped in the gauge and the term effective dead volume is often used to
express the fact that a gauge is not a rigid part of the gas pipeline but will
increase in volume at high pressure and decrease at low pressure. These
changes in volume can, if linear in pressure, be expressed in terms of an
equivalent volume of gas.
Effective dead volume leads to inaccuracy in the measurement of pressure
18
Distinction should always be maintained between gauge pressure, measured relative to
atmosphere, and absolute pressure, measured relative to vacuum. Care should be taken
always to express only di fferent i al pressures in units like bar or psi, and always to specify
single-ended pressures as bara or barg, psia or psig.
in small isolated systems at high pressure. Real dead volume gives this effect
but also, more seriously, acts as a reservoir of gas which is not swept by flow.
In ultra-pure systems, or in systems handling toxic or explosive gases which
must be purged, dead volume is highly undesirable.
In extreme cases, the gauge can be filled with a hydraulic oil, such as
silicone oil, reducing the real dead volume, although some effective dead
volume will remain. Precautions should be taken if oil is used, however, that
contamination of the gas line is avoided. Also, some gases will dissolve in the
oil, raising the dead volume again. Some gauges avoid the latter two
problems by retaining the oil behind a thin diaphragm or membrane.
The other major defect of the Bourdon tube is the hysteresis intrinsic to
the mechanical gearing system applied to the tube end.
For low-pressure work, the Bourdon tube motor for the mechanical
pressure gauge is replaced by an aneroid capsule. This achieves greater
sensitivity as it comprises one or more thin diaphragms with either a pressure
connection or a vacuum inside. The standard domestic barometer is an
example of the latter.
Manometers. Manometers using liquid in U-tubes are now used rarely in
industry. U-tube manometers are nevertheless still used in laboratories,
although the use of mercury as the operating fluid, whose high density makes
higher pressures more convenient to measure, has been discouraged by
strict health and safety regulations. U-tubes offer simplicity of operation,
potentially high accuracy and the minimum of calibration. However, it is
usually necessary to use glass sight tubes, with the problems above, and they
are quite difficult to read accurately. By sloping the U-tube, and using a
low-density liquid such as paraffin, U-tubes can be used for small pressures.
These low-reading instruments have now largely been supplanted by digital
readout electronic pressure sensors, now available at relatively low cost.
Electronic pressure sensors. Electronic pressure gauges have now
become ubiquitous in industry and are gradually replacing many other types
of gauge. The basis of a number of related electronic pressure gauges is the
electronic sensing of a small movement of a robust stainless steel diaphragm.
Capacitance gauges are the most sensitive, using differential capacitance
sensing. They can work anywhere from near vacuum (pressures of bar)
to hundreds of barg, while differential transformers are satisfactory at
higher pressures. Capacitance cells are capable of the highest accuracy and
are used in equipment where precise pressure is critical. Processes which
employ subatmospheric, but not high vacuum, pressures often use capaci-
tance cell sensors. They have supplanted some of the vacuum gauges such as
the Pirani gauge where accuracy is needed. At the very highest pressures
another form of sensor becomes possible where capacitance change sensed
is intrinsic and not due to the movement of a diaphragm: one in which the
GAS TECHNOLOGY 161
small change of capacitance of a capacitor with pressure is employed
(Andeen, 1971).
Strain gauge detection of the diaphragm deflection leads to a lower cost
sensor but one in which drift and noise limit the lower limit of pressure
compared to capacitance cells. Electronic pressure capsules can be designed
so that the sensing diaphragm is right at the edge of the sensor. With this
design electronic pressure gauges effectively achieve much lower dead
volumes than any mechanical or manometer gauge can possibly achieve.
There are now tiny silicon-strain-gauge-based pressure sensors. These
integrate minute piezoresistive silicon resistors connected in a bridge format
with a silicon diaphragm etched out of a piece of silicon a few millimetres
across. The silicon is mounted so that gas pressure bells out the diaphragm
slightly and differentially changes the piezoresistance values. A simple
temperature-compensating and amplifying circuit completes the sensor.
These sensors can be made very cheaply; the major cost is often the
calibration. (They can even be made disposable; for example, disposable
sensors are used to sense dynamic blood pressure pulses in heart surgery.)
Again, at higher pressures, it becomes possible to use electrical resistance
change in an intrinsic sensor. Alloys such as manganin, originally destined
for use as a low-temperature coefficient resistor, can be used in a sensitive
bridge circuit to sense pressure via their resistance change with pressure
(Anon, 1975).
Temperature measurement. An important element in many measurements
on gases is their temperature. This is obviously needed when using
volumetric flow measures, for example, but is also needed for corrections in
mass flow measurement and is useful in monitoring cryogenic liquids.
At normal temperatures any of the normal or elevated temperature
measurements will serve, with convenience, reliability and cost being the
main considerations. The small thermal capacity of low-pressure gases
occasionally leads to long time constants or errors due to conduction along
the thermometer supports, and this is something that needs attention
occasionally.
At low temperatures there are more problems. Liquid-in-glass ther-
mometers using mercury are unworkable as they freeze, although coloured
thermometers based on coloured ethanol (the lowest freezing of the
common alcohols) freeze only at 117C. Solvent mixtures based on
liquid-in-glass thermometers are available down to 120C (153 K). Liquids
with large liquid range, such as ethanol (117 to 78C), isopentane (160 to
28C) and isopentanol (117 to 131C) are the most useful.
19
19
Interestingly, ethane (freezing point 90 K) was used at one time, in combination perhaps with
other hydrocarbons and a dissolved blue dye, as the active liquid in liquid-in-glass ther-
mometers down to cryogenic temperatures.
Wirewound resistance thermometers are generally bulky but they are
simple and function at all temperatures. Cheaper are carbon film resistors,
of the type used in electronic equipment. Because of small changes in
thermal cycling down to cryogenic temperatures, these are not suitable for
accurate work. However, they work right down to liquid helium tempera-
tures from room temperature and are easily calibrated, providing a
relatively fast change of resistance with temperature down to the lowest
temperatures. Resistivity of metals, such as high purity copper, follow a
steady fall in resistance, although below 10 or 20 K or so, depending on the
degree of impurities in the metal, the resistivity changes more slowly with
temperature.
Thermocouples continue to function to the lowest temperatures but the
signal obtained at cryogenic temperatures is small and noisy. A gold/2%
cobalt alloy on a copper thermocouple has over twice the thermoelectric
power as the standard copper/constantan thermocouple (around
at 80 K. Compared to thermocouples the signal available from simple diode
thermometers is bigger and less noisy. A simple silicon diode basically has
forward voltage drop d which is proportional to absolute temperature at a
constant current, and diodes specially trimmed are now available which give
precise temperatures with large, easily used electrical signals. Once these
simple and inexpensive devices become as well-known to mechanical and
chemical engineers as they are to electronic engineers and physicists, their
use will become much more widespread.
2.3.3 Analysing gases
The purity of gases is an important part of their specification. For example,
in an application where argon is used in a high-temperature furnace to
protect against oxidation or nitridation by the air, the presence of or
will be very important. Also, the monitoring of gas composition in
production plant and gas-using plant is a useful diagnostic for plant
problems. For example, the presence of additional helium or neon in
nitrogen from an air-separation plant might indicate failure of the column
top vent. The presence of hydrogen in the nitrogen from a PSA plus purifier
nitrogen supply might indicate failure of the catalyst in the purifier. After a
failure incident involving gases, it is important to analyse the gases used for
purity and correct majority component. Occasionally inexplicable failures
are traced to the wrong gas having been put into a cylinder or to residual gas
in a cylinder which was not blown-down. For these reasons, analysing gases
is an important part of industrial gases operations and all gases are supplied
to an analytical specification with, if required, a certificate of analysis (Table
2.8).
The analysis of gases for their molecular constituents is itself a large
subject, with many books and journals devoted specifically to it. As early as
GAS TECHNOLOGY 163
the late eighteenth century gas analysis had become well-developed, so that
Lavoisier could remark in his chemistry book of 1789: It is necessary to
become familiar with the analysis of gases by long experience. His book in
fact supplanted that need by describing most of what is necessary to carry out
bench analyses of gases. Gas analysis is now a very powerful branch of
chemical analysis, so much so that many analyses are carried out by
converting samples into gas so that the power of sophisticated gas analysers
can be turned on to them. This section deals with analysers for industrial
gases, however, and with continuous reading instruments, rather than with
bench methods and single batch measurements.
Gas analysers can be complex computer-driven instruments capable of
simultaneously analysing hundreds of gas components. Multigas analysers,
such as gas chromatographs (GC) and mass spectrometers (MS), are used
routinely in industry. However, the workhorses of the gases industry are
undoubtedly single gas analysers of various types. Most of these have now
attained a certain level of sophistication and are supplied by specialist
suppliers. These are described in later sections. However, the simplest of
these are really exceptionally simple, very low in cost and can often be
constructed by the user rather than by a specialist. Some examples of simple
analysers actually in use in particular circumstances are as follows.
Two electrodes in a distilled water bubbler (for acid gases such as and
perhaps also for alkalis such as ammonia). With precautions to
ensure good dissolution, the conductivity of the water gives a precise
measure of the gas which has been bubbled through. This principle, with
the addition of hydrogen peroxide to the water to convert and to
sulphuric and nitric acids, is used in the statutory measurement of acid gas
emissions from exhaust stacks, for example.
A hot wire which, when it is exposed to carbon-containing gas, changes its

resistance. A number of common resistive wires show this effect, which is
sometimes seen as a nuisance in katharometers used in gas chromato-
graphs. At one time this effect was used in instruments indicating furnace
atmosphere in metal heat treatment or carburisation. The wire in effect
was being heat-treated and carburised, too, its electrical as well as its
mechanical properties being modified.
A dual hot wire cell, the katharometer used for hydrogen detection. The
katharometer simply measures the thermal conductivity of a gas and in its
simplest form comprises a bridge circuit with hot filaments in each arm of
the bridge. One filament samples the unknown gas, the other a known
zero gas.
Absorption and titration is always an option for many gases and has the
same advantage as the refractometer of absolute calibration. can be
estimated with alkalis in this way, oxygen by absorption in pyrogallol and
hydrogen by reaction with palladium chloride followed by titration of the
hydrochloric acid produced.
The philosophy generally followed in gas installations has been to install a
mi ni mum number of on-line single gas instruments of various types. Where
these are reliable enough, they can be used both to observe and to control
the process on which they are used. The on-line single gas analysers are
backed up with occasional off-line multigas analysis, or occasionally by an
on-line multigas instrument. The multigas instruments are typically used to
check the function of the single gas units and to check for unknown
contaminants.
2.3.4 Gas chromatographs
GC is probably the most powerful single technique available to the analyst.
Although GCs are most often seen in the laboratory and used for off-line
analysis, an automated GC technique is also the basis for many on-line
automatic analysers, although this is often not obvious from their external
experience. Automated GC analysers, although complex, may run reliably
for years with little intervention. GCs for analysing the contents of
argon-rich streams in ASUs and GCs measuring trace and CO in inert
gases are examples of on-line use.
The GC concept is straightforward. A steady stream, typically of pure gas,
the carrier gas or mobile phase, is passed along a long thin tube, the
column. The tube is packed with a solid absorbent, the stationary phase,
such that the gas may still easily pass. The (small) sample for analysis is then
injected swiftly as a bolus into the gas stream. The components of the
sample bolus are then reversibly absorbed by the first part of the column,
then released and carried a little further up the column before being
absorbed again (Figure 2.34).
In this way the sample components proceed along the column, progress-
ing more slowly than the carrier gas. Their speed of progress depends on the
carrier flow rate and on how much the equilibrium absorption of the
component favours the stationary phase. If the component is highly
absorbed, then progress is slow; if it is only slightly absorbed, then it is fast.
In normal practice the carrier gas should not be one of the components to
be measured. Helium, hydrogen, nitrogen and air are probably the
GAS TECHNOLOGY 165
Figure 2.34 Separation of a mi xt ure in a GC column showing typical trace characteristics.
commonest carriers, being chosen in many cases to be the best match to the
detector employed, and on economical considerations. The carrier gas
should at least be free from troublesome impurities like water (which slowly
degrades molecular sieve columns), and should ideally be of the highest
purity grade conveniently available. For ordinary work, a standard cylinder
gas, perhaps with a molecular sieve trap to remove moisture, is adequate.
For ultra-trace work (below 1 ppm) small purifiers (for example, metal
getter-types for Ar/He or a palladium diffuser for should be considered.
The stationary phase may be an inert solid support coated with a thin layer
of a low vapour pressure liquid as the active absorbing component. More
common, however, are solid microporous stationary phases such as the
zeolites or Porapak.
The result of all this, if the carrier gas and stationary phase have been
correctly selected, is to separate the sample components so that they come
out (elute) from the column separately. A simple non-selective detector
can then be used and the area (integral of concentration with respect to
time) of the peaks detected measured to give the relative concentrations of
the components.
The standard laboratory gas chromatograph employs a high-quality
polymer septum and samples are literally injected with a calibrated syringe
through the septum using a needle. However, for on-line GC, and for most
industrial routine work, it is simpler and more satisfactory to use a sample
loop on a multiport rotary valve for injecting a bolus of sample into the GC
column. The loop is filled with flowing sample when switched into the
sample position, the valve is rotated to cut off the gas sample and the sample
then placed in the carrier gas path (Figure 2.35).
For use as an on-line gas analyser the GC must be automated. Such
Figure 2.35 Complete GC gas analysis system.
automated or process GCs have been used for many years now and have
been substantially perfected. Servo-driven or solenoid valves automatically
take samples and carry out any other switching. Resultant peaks are
integrated and the results fed either to a digital interface for display on a
computer screen or to digital-to-analog converters (DACs) to give a voltage
level for each peak, i.e. acting as a pseudo-single gas analyser output.
Routine maintenance is confined to occasional replacement of columns and
checks on calibration. A typical automated GC application would be the
analysis of crude argon streams (containing nitrogen, oxygen, argon and
other rare gases) in an air-distillation plant. The process GCs used here are
relatively simple systems with a single isothermal room-temperature
operation molecular sieve column and a thermal conductivity detector using
a helium carrier gas.
GC column choice. By far the most popular stationary phases in GCs for
gas analysis are microporous minerals, such as zeolites, or microporous
polymers, such as the divinylbenzene copolymer-based Porapak
TM
ma-
terials. These are packed into long tubes of 1/8" or 1/4" diameter (2 mm
5 mm or so internal diameter) clean stainless steel, up to 2 or 3 m long, which
are then wound into a helical coil if required. It is an advantage to use the
smallest column which can be reasonably easily handled and packed. A
small column will usually yield the best resolution and is of low thermal mass
so that temperature ramping during analysis is possible.
It is possible to use non-packed columns, popular in chemical GC work, in
gas analysis. However, these need to be of small internal bore (less than
0.5 mm or so) and considerable length (50 m is typical). Newer configur-
ations include the PLOT (porous layer open tubular) columns with a porous
GAS TECHNOLOGY 167
Figure 2.36 Separation of a gas mixture on a Porapak column.
alumina coating on the inside of the glass or quartz capillary tube. The
porous alumina can be further coated with another microporous material if
necessary. PLOT columns are used commercially in light hydrocarbon
separation.
Molecular sieve (zeolite) columns usually employ type 5 (5 pore size
aluminosilicate) or 13X (10 aluminosilicate) granules in a 1/4" column.
These must either be purchased pre-activated or must be activated before
use by driving off water and CO
2
, both of which molecules will sit in the
active absorption sites of the sieve and prevent correct separation of other
gases. The activation is conveniently carried out by heating to above 400C
with dry carrier gas flowing. Molecular sieve columns are used where
separation of air gases is required. However, separation of argon from
oxygen is difficult, requiring cooling of the column below ambient, and CO
2
tends to absorb and tail badly. Porapak columns use polymer beads, mostly
prepared from copolymers with divinylbenzene (Figure 2.36). Chromo-
sorb is another trademarked commercial material with similar properties.
They provide good separation of most gases, with the exception of the air
gases, although air gases are nicely separated at low temperatures on
divinylbenzene copolymers. can be analysed with these columns, as can
water vapour with difficulty (tailing tendency again).
Problems with gas chromatography. Problems with gas chromatography
are frequent but it is a measure of the versatility of the technique that they
can often be overcome. For example, take the problem of sample size. With
most analytical techniques, the smallness of the samples needed is a measure
of the efficiency of the technique. However, in GC the sample size analysed
is typically only a millilitre or less, which is an advantage that often gives a
problem. In many situations the gas is inhomogeneous to some extent. It is
therefore necessary to ensure that the tiny sample taken is truly representa-
tive of the mixture. An obvious example of this is where a liquefied gas is
being sampled: the gas phase may well have a different composition to the
liquid phase and samples must be pure liquid or pure gas if the results are to
be quantitative.
High boiling point vapours admitted to the chromatograph (for example,
oil in compressed gas or solvent in acetylene) will take a long time to exit the
column. A zeolite with a retention time of just a minute for nitrogen might
have a retention time of hours for a vapour, which would make laboratory
GC tedious and process GC impossible. This problem is also present with
polar molecules such as water, which in addition tend to cause tailing after
their peaks. The longer retention times can be compressed closer in time by
ramping the column temperature after sample admission as retention times
rapidly decrease at high temperatures. High boiling point vapours can also
be dealt with by back-flushing. After 1 or 2 min for analysis of the desired
light components, instead of waiting hours for the heavier components to
elute forwards down the bulk of the column, the column flow is reversed for
2 or 3 min, eluting the heavy components swiftly backwards from their point
of progress near the inlet.
A GC column can be overloaded with the majority component of a
mixture in which it is required to measure a minority component or trace
gas, which is the sort of problem which is general to any analytical technique.
Unlike many other techniques, however, in GC the problem can be dealt
with by using two columns. The first column is arranged to feed the second
column via a divert valve. During the peak of the majority component, the
first column effluent is diverted to waste. The second or main column can
then be used for analysis with most of the majority component absent,
enabling accurate work even with trace gases.
Some gas mixtures are only partially separated by any single GC
column/carrier gas combination. By passing an unseparated peak through a
second column of a different material (so-called heart-cutting), the
components can be completely separated. In this way, the problem of
incomplete analysis of a particular mixture by any one GC column is soluble;
provided there are no components which have identical retention times on
all column materials, then any mixture can be separated.
Finally, the GC technique is limited by obvious technical factors: it is a
very sensitive analytical tool and requires clean systems and consumables.
Impurities in the carrier gas, for example, can give rise to a number of
problems such as baseline drift, noise and reduced sensitivity. The injection
septum or valve needs care to ensure that bleed (continuous baseline due to
compounds deposited on the inlet slowly evaporating, outgassing or
decomposing) is avoided.
Gas chromatograph detectors. Since the GC column should have separ-
ated the sample mixture into peaks separate in time, the only essential
requirement for a GC detector is that it distinguishes gas containing any
sample component from the carrier gas.
GAS TECHNOLOGY 169
To be a good GC detector, however, a number of other qualities are
highly desirable:
linearity: this means that peak area can be used to measure components
quantitatively;
high speed: to achieve a reasonable analysis cycle time on mixtures of

components with widely varying elution speed, fast-eluting components
will tend to elute in just a few seconds;
large dynamic range: with mixtures of widely varying elution speed, high
concentrations of components with a fast peak require a good high-level
measurement, whilst low concentrations of slow components need high
sensitivity;
drift free: slowly eluting components, with peak widths of up to several

minutes, will be confused with background drift at lower concentrations,
if drift is large.
Variations in carrier gas pressure, flow, temperature and composition will
all produce baseline drift. Different detectors are sensitive to different
carrier gas changes. For example, flame ionisation detectors are sensitive to
H
2
concentration in the flame, whilst flow variations cause drift with
katharometers.
The most common GC detectors are as follows.
TCD (thermal conductivity detector or katharometer). A TCD simply
passes the column effluent over a hot wire or hot thermistor. The cooling
effect of the gas varies a little when it contains some eluted sample, and this
varies the temperature and therefore resistance of the filament. This change
of resistance is recorded as the signal. In practice, because the resistance
change is small, two or four detectors are used, with the resistance of pure
carrier gas cells being compared to the resistance of the column effluent cells
and additional reference cells out of any gas flow as well if necessary. The
dynamic range of the TCD is poorer than the FID, and it is around a
thousand times less sensitive. Despite this, it is the most commonly used
detector. The possibility of an inert carrier gas such as He, or Ar and the
complete universality of the detection principle means that it is a very
attractive detector. Table 2.9 clearly shows the ready detectability of most
gases with a helium carrier and shows that for many gases of interest a similar
response is seen, minimising calibration problems. (Note that hydrogen
gives a peak in the opposite sense to the other gases, which can give
problems with amplifiers and using a chart recording efficiently. Also the
peak is nonlinear with concentration, unlike the other gases.)
FID (flame ionisation detector). A simple FID takes carrier gas
from the column and burns it in pure air, perhaps adding an additional fuel
gas stream, measuring the current flowing through the flame between two
Figure 2.37 Flame ionisation detector for GC.
plates with a differential voltage, up to 100 V or so, across them (Figure
2.37).
20
It is highly sensitive with a high dynamic range and a response which
is almost exactly proportional to the concentration of C atoms, which ionise
in the flame, in the carrier gas stream. However, it does not detect
non-carbon gases (or indeed with usable efficiency, severely restricting
its usefulness, and needs a very sensitive amplifier, since the ion current is
typically only picoamps or nanoamps.
For detection of low levels of it would be useful to be able to use the
simple but sensitive FID detector. This can be achieved with the expedient
of a methaniser unit. A very simple heated (300400C) catalyst tube,
containing nickel on an inert microporous mineral support, placed after the
20
In many ways the FID is the simplest of all detectors to construct and operate; the author
made a successful FID device while at school from pieces of wire and laboratory glassware.
GAS TECHNOLOGY 171
hydrogen mixer but before the flame results in almost 100% conversion of
to which is then detected in the FID.
The use of additional fuel gas means that the FID flame will not extinguish
if a large bolus of non-inflammable gas appears and means that a greater
dynamic range can be accommodated by the detector. The fuel often used in
FIDs can be pure hydrogen but a mixture of 40 or 50% hydrogen with
nitrogen or helium is a useful alternative. Similarly, FID GCs can also be
operated with carrier gases other than hydrogen; for example, helium,
nitrogen or even air. (If air is used, however, it must be free of
hydrocarbons; typically a purifier unit or a cylinder of compressed pure air
will be required.)
ECD (electron capture detector). A Nickel-63 -radiation source ionises
the column effluent, or Ar carrier gas often with methane added, whose
electrical conductivity is measured. This detector is sensitive to electrophilic
molecules such as halogenated compounds. These compounds reduce the
standing current in the conductivity cell, rather than increasing it, and
because of this the ECD detector is restricted in its dynamic range.
Ultrasonic detectors. By detecting the change in the speed of sound in
the effluent gas, many gases can be non-selectively detected against an or
He carrier as these carriers have a very high sound velocity which drops
sharply with heavier gases present; the velocity of sound is inversely
proportional to the square root of mean molecular mass. Although
ultrasonic detectors are not particularly sensitive, they can be made very
robust and can be protected against agressive chemicals and so are often
used in the analysis of toxic and corrosive cylinder gases.
In addition to these and other non-selective detectors which are the
workhorses of GC detection, most selective gas analysers, such as those
described below, can and are used as GC detectors. An example is the flame
photometric detector, used for detecting and It uses an optical
filter to look at a hydrogen flame to which the sample gas is added, detecting
sulphur atoms via their 394 nm optical emission (phosphorus atoms also
emit, at 526 nm) to a photomultiplier.
Conversely, most non-selective detectors normally used with GC can be
used as single gas analysers, where the situation confines the gases measured
so that no confusing results can be obtained. An example of this is the gas
sniffer used to detect gas leaks in the domestic gas distribution network.
These sniffers are sometimes mounted on the front of a van. As the vehicle is
driven along the road at low speed, gases escaping from the ground near a
leak are detected by a flame ionisation detector, alerting the crew to stop and
investigate. Another example is oxygen depletion analysis in air using
devices based on an ultrasound principle. These devices make the assump-
tion that any increase in sound velocity is due to a decreasing amount of
oxygen and sound an alarm at, say, 15% oxygen content on this assumption.
The user must ensure that in the situation where they are used there is no
possibility of a heavy gas accumulating in the atmosphere to an extent
sufficient to mask appreciably the decrease in oxygen content registered.
GC integrators. It is invariably possible to connect a laboratory GC to a
chart recorder of some description, plotting detector signal versus time. In
many cases, a GC trace on a chart is all that is required to indicate
approximate concentrations of different components in a mixture. How-
ever, it is difficult to estimate concentrations accurately especially if (as is
typical) components have radically different peak widths and if peaks are of
radically different areas.
The skilled operator can obviate some of the latter effect by changing the
chart recorder gain control and chart speed at the appropriate moment.
However, the use of computer-based electronic integrators is now universal.
These provide a wide dynamic range, integrate accurately over peaks, no
matter their width or if they are asymmetrical, apply calibration factors and
give out their answers in convenient form; for example, lists of numbers
giving gas concentrations. In the laboratory, it is possible to use a desktop
computer, such as a PC compatible, as the integrator, although stand-alone
integrators are favoured for convenience and robustness in industrial on-line
situations.
For small peaks, and for chromatograms where the baseline signal from
the carrier gas drifts, it is important to use an integrator with sophisticated
peak detection and baseline correction. A typical algorithm declares a peak
to have started when the slope of the signal versus time exceeds a threshold,
and stopped when the negative slope value on the return to baseline falls
below the threshold. A baseline correction to be subtracted from the peak
area can be made by calculating the area of the trapezoid under the peak
made by the line from the start value of the peak to the stop value.
Corrections can also be applied for tailing and for separating peaks which
overlap to some extent (see Figure 2.34).
It is still a wise precaution to run off a chart picture of the chromatogram,
at least occasionally, because these sophisticated integrator algorithms can
sometimes fai l . Most integrators can help in the production of a useable
chart recording trace by changing amplifier gain factor and timebase
automatically at pre-programmed appropriate moments, or by printing off a
record simulating this from memory.
2.3.5 Infrared and related analysers
Simple diatomic and monatomic gases are completely transparent to
infrared (IR) radiation. However, most other gases are selective infrared
absorbers, with complex bands of absorbtion and transmission. The science
GAS TECHNOLOGY 173
Figure 2.38 Rotational and vibrational structure in infrared absorbtion spectra. (a) Coarse
resolution vibration spectrum; (b) fine resolution vibration/rotation spectrum.
of infrared absorbtion by gases is complex but basically the absorbtion is due
to the IR radiation exciting vibration (or vibration/rotation) modes in the
molecules (Figure 2.38). The frequencies at which vibration occurs depend
on the sizes of the masses and springs at the molecular scale, the atomic
masses and bond strengths, and are largely characteristic of a particular
compounds. The vibration spectrum of each gas has typically a complex
comb-like structure, with many sharp peaks of differing heights in the region
from 3 to 20 The fine structure reflects rotation excitations, whilst the
envelope within which the sharp peaks lie relates to vibrational energy
levels.
This fact is employed in two types of infrared analyser: the non-dispersive
and the dispersive IR analyser, which measure IR absorbtion at many and at
a single frequency, respectively. The non-dispersive types are popular as
on-line single-gas analysers, whilst the dispersive types are mainly used in
the laboratory.
Non-dispersive IR analysers (NDIRs). NDIR as a principle was developed
rapidly as soon as good, reliable electronics were available for the
capacitative detectors used in the early Luft analysers. There are many
different designs but all are based on a similar set of principles (Figure 2.39).
Figure 2.39 NDIR analyser principle.
First an infrared beam is created, typically by filtering out the visible
light from a filament lamp or perhaps from a low-temperature glower
source. In addition, mid-infrared LEDs and laser emitters which offer
improved performance are available today. The beam is then passed
through a filter wheel comprising a dummy reference cell and a filter
cell, which will be a cell full of the gas to be detected. The resultant
infrared beam is passed through a coarse infrared filter and then the sample
of gas to be measured. Lastly, the beam is absorbed and converted to an
electrical signal by means of an infrared photoconductive (lead selenide is
typical) or pyroelectric detector.
The filter cell filters out all the radiation lying in the multiple narrow
bands to which the gas detected is absorptive. Hence, if any of this gas is
present in the sample then the unfiltered detector output will fall while the
filtered detector output remains constant. The coarse IR filter makes sure
that the band of IR used through the sample cell is identical. The
difference signal between the two is proportional to the gas concentration
in the sample (at least at small concentrations). The single detector is
presented first with the filtered and then with the unfiltered beam in
sequence. The a.c. signal produced at the detector is proportional to the
gas concentration.
Although convenient, NDIR instruments are nonlinear, sensitive to
contamination, have a tendency to drift and are not perfectly selective for
the gas to which the filter cell sensitises them. They are normally not used
for very high sensivity (below 10 or 100 ppm, say) but can be made very
sensitive (ppm level or below) by using long-path, multiple-reflection gas
cells. Common instruments analyse for CO, and Car
exhaust emissions are now tested in most countries of the world for CO and
hydrocarbon content using NDIR analysers. Similarly, most hospitals have
in everyday use capnometers using NDIR for measuring a patients
breathing via the metabolic coming off the lungs.
IR spectrophotometers. Although normally only used in off-line mode in
the laboratory, IR spectrophotometers are important instruments for
measurements on unknown mixtures. A wavelength-selected IR beam is
first made by means of a diffraction grating. The output from this
monochromator is then passed through the unknown gas mixture and the
beam detected with a solid-state IR detector. The output as a function of
the monochromator wavelength, the spectrum, can then be compared with
library spectra of gases.
A qualitative look at what gases are present can often be seen by eye
whilst with some computer deconvolution work quantitative results can be
obtained.
An alternative form of IR spectrophotometer, Fourier transform IR, uses
an interferometer type of set-up, which produces the Fourier transform of
GAS TECHNOLOGY 175
the infrared spectrum. This output can then be unravelled into a spectrum and
gas concentrations with a computer.
Raman spectroscopy. Until the advent of powerful and inexpensive lasers,
Raman spectroscopy for any practical purpose was an impossible dream.
Even today, special techniques such as operating inside the cavity of the laser
(so-called active cavity working) are often necessary to make Raman useful.
The Raman effect is the small amount of light scattered off atoms that is
shifted in frequency by an amount due to excitation of a molecular vibration.
Spectral lines with shifts downwards in frequency are known as Stokes lines,
whilst lines shifted upwards in frequency are known as anti-Stokes. Using a
powerful laser, and rigorously excluding any of the scattered light which is
unshifted in frequency as well as the laser beam itself, it is possible to see a
unique spectrum for diatomic as well for polyatomic gases. For nitrogen, for
example, about 1 part in 10 million of the 488 nm light from a green argon-ion
laser is shifted to 550 nm. The spectra for polyatomic gases have strong
similarities to the infrared spectra, relating as they do to molecular vibration.
Basically, the energy levels in the molecule are the same but the transitions
allowed between energy levels differ.
One of the first gas analysis applications tried for Raman spectroscopy was
the monitoring of pollution plumes. Very powerful blue or green lasers are
arranged to illuminate plumes from power stations or other industrial
complexes. Via large aperture coaxial optics, light back-scattered from the
plume, several hundred metres up in the air, can be analysed for Raman
spectrum.
21
21
As recently as 1984, the author was witness to what may well have been one of the first routine
practical applications of the technique to gas analysis: the monitoring of oxygen content inside
ampoules of a sensitive medical drug sealed under nitrogen. A powerful argon-ion laser was
shone through the gas space at the top of the ampoule and the ampoule rejected if more than a
pre-set amount of oxygen Raman line was seen using sensitive photomultiplier detectors.
Today the technique is still a difficult one; the laser used in ampoule monitoring was 2 m long
and weighed over 200 kg. Its sensitivity to gas concentrations is not high, parts per million
monitoring is very difficult, and its commercial value is still not clear. Nevertheless the technique
is today more widespread. A Raman-based anaesthetic gases monitor has been developed.
Following laboratory work with an argon-ion laser, a commercial instrument is now available
using a HeNe laser which will monitor six different gases at the fractional percentage level
simultaneously, rivalling mass spectrometric techniques (section 2.3.6) in the percentage range.
There are still problems with such analysers, however. In order to raise their sensitivity to the
highest possible level, the gas must be led inside the laser cavity where the optical electromagnetic
field is at its maxi mum. This means that small deposits from the gas being analysed on the optics,
or even too large a refractive index difference between the air used to protect the optics and the
sampled gas, can lead to the laser cavity gain varying, causing problems. In addition, despite these
efforts, signals are still very weak, even on gas concentrations in the fractional percent range.
Finally, in order to use HeNe lasers tried-and-tested (and very sensitive) photomultiplier
detectors must be replaced by the more di fficult technology of avalanche photodiode detectors.
The latter are diodes with an intrinsic gain effect, which have a good sensitivity for the Stokes
Raman lines, which are in the near IR, being 30 to 150 nm longer in wavelength than the 633 nm
HeNe excitation.
Optical interferometric refractometry. One gas detector that is often
neglected is the refractometer. The gas refractometer is based on an optical
interferometer which is the only ready means to detect the small differences in
optical index between gases at ordinary pressures. A refractometer has the
advantage over most analysers in that its response can be calculated from first
principles and can be used as a standard to calibrate other analysers. The UK
National Physical Laboratory and US Bureau of Standards, amongst others,
carry out these kind of measurements in standards work, although
commercial instruments are now available. The availability of inexpensive
laser light sources and electronics may make refractometers more attractive
than in the past, although simpler optical components and large-scale series
manufacture will be needed to reduce costs to the levels seen in many other
analysers.
2.3.6 Mass spectrometry
The mass spectrometer (MS) is often a sophisticated and expensive analyser
and might not be thought of as useful in industrial gas analysis. However,
simple MS devices can cost as little as a few thousand dollars and MS offers
some unique advantages, being a technique almost as versatile as GC.
The basis of MS is to take a minute sample gas flow and leak it into a
low-vacuum region (the source) where an electron beam of a few electron
volts energy gently breaks the molecules into molecular and submolecular
ions. These ions are then sucked out of the source region by high voltage plates
with small holes in them. The accelerated ions then pass through the holes into
a region of high vacuum mbar or less) where a dipole d.c. magnetic
field or a quadrupole a.c./d.c. electric field is used to select them by
mass/charge ratio. Only ions of the mass/charge ratio selected by the magnetic
or electric fields will penetrate as far as the exit slit. After the exit slit a detector
measures the flux of ions. A computer drives the magnetic or electric fields,
effectively scanning the mass/charge spectrum of the ions coming from the
source region.
The mass spectrometer is, like the gas chromatograph, fundamentally a
means of separating molecules rather than a detector per se. However, in the
case of the mass spectrometer the separation is achieved on charged
molecules, molecular ions, which constitute an electrical current flow in
themselves. The simplest and most common MS detector, therefore, is a
metal plate or cup (Faraday cup) connected directly to a sensitive
electrometer amplifier.
For measurements at very low concentrations, ppm or below, an electron
multiplier detector becomes useful. An ion incident on the multiplier causes
an avalanche of electrons to be released. The resultant large current is
simpler to detect than the tiny molecular ion current. The most common
multipliers consist of a highly resistive coating on a curled horn-shaped
GAS TECHNOLOGY 177
Figure 2.40 Quadrupole mass filter.
insulating tube. For ultra-trace detection, individual electron avalanches are
counted; as little as one molecule can therefore be detected.
The highest mass resolutions are obtained by magnetic mass selection
instruments. Large magnetic sectors can achieve resolutions of 1:10 000 or
more, which means that nominally isomass ions (those with the same number
of protons and neutrons, such as and CO, for example) can be
distinguished. Such resolution can greatly simplify the interpretation of mass
spectra but is expensive to achieve. In almost all gas analysis work the
maximum resolution needed is that to distinguish ions differing by at least 1
atomic mass unit. With a maximum useful molecular ion weight of around 200
or 300, resolutions of 500 or so are entirely adequate for this.
The cost of mass spectrometers has been reduced considerably in recent
years with the advent of ~ 500 mass resolution quadrupole instruments and
inexpensive electronics and computer control. The quadrupole mass
spectrometer principle is that of a molecular mass/charge ratio filter. Four
accurately ground parallel cylinders are charged both with d.c. and a.c. (at a
few MHz) to create quadrupole electric fields and ions are admitted along the
central axis of the quadrupole field. Figure 2.40 shows the quadrupole
arrangement. Heavy ions tend to oscillate in the a.c. field with increasingly
large amplitude along one plane between the cylinders, while light ions tend
to oscillate in the orthogonal plane. Ions of the tuned mass, however, pass
through the assembly unaffected. The ratio of d.c. to a.c. electric field
determines the mass/charge ratio passed.
The mass spectrometer is completed by the addition of digital-to-analog
converters (DACs) and analog-to-digital converters (ADCs) driven by a
computer. The computer is programmed to admit gas samples, produce
Figure 2.41 Simple quadrupole mass spectrometer.
ramp or staircase voltages to scan mass/charge ratio and record the resultant
mass spectrum via an ADC in digital form. The mass spectrum can be
displayed directly, output via DACs as a number of pseudo-single analyser
voltage signals, or further processed via calibration factors and even
mathematical (matrix) processing to yield gas concentrations. The use of
matrix inversion procedures often enables the analysis of gases whose
mass/charge ratio peaks overlap, provided the mixture components are
known. Figure 2.41 shows a complete quadrupole mass spectrometer
system.
Despite the low cost of the quadrupole MS, complete MS instruments are
still relatively expensive at $15 000 or so. Much of the cost of this kind of
mass spectrometer can now be attributed to the requirement to work at high
vacuum. Even the smallest instrument needs high vacuum pumping
equipment costing many thousands of dollars.
Problems with mass spectrometry. Just as for gas chromatography, there
are limitations with mass spectrometry.
Some gases have virtually identical mass/charge ratio molecular ions. CO

and are the most common examples.
Ion sources are more or less non-selective. This means that although
calibration is simpler, with factors of order unity, it is difficult to measure
trace components in the presence of main components. Selective ion
sources such as photoionisation have mostly proved to have too many
other problems to be useful, although atmospheric pressure silent
discharge ionisation is now commercially available and very usefully
selective.
GAS TECHNOLOGY 179
The sample pressure must be reduced from its source pressure to about
mbar or so. This can be done by a simple leak but such an
arrangement leads to a very slow response. A solution to this is to reduce
pressure in two stages: a leak down to mbar, typically via a capillary
tube, and then via porous ceramic leak from to mbar. Most of
the sample is thus bypassed to the vacuum pump pumping the intermedi-
ate volume.
Some sample gases will react with the surface of the filament in the ion
source, changing the electron beam current and ultimately destroying it.
Oxygen reacts with residual carbon in tungsten filaments, producing CO

gas in minute amounts sufficient to make mass 28 measurements
unreliable in the presence of oxygen. The use of rhenium filaments
alleviates this problem.
Atmospheric pressure ionisation MS (APIMS). APIMS is a technique
which has only in the last five years become commercialised. It is a
specialised technique answering the very specialist requirements for
measurement of ppb level trace impurities in the bulk inert gases nitrogen
and argon, which have come from the ultrapure gases needs of the
semiconductor industry.
In its simplest form APIMS involves sucking the sample gas away from a
silent discharge needle ionising region and stripping away the bulk of the
uncharged gas molecules whilst focusing the small numbers of ions along the
axis of a quadrupole mass analyser (Figure 2.42). In practice, this is done by
two or more skimmer cones. These are metal cones with a small (100 or
so) central aperture, insulated so that suitable d.c. electric potentials can be
applied. Vacuum pumps reduce the pressure behind each cone until the
mbar or less needed to operate a quadrupole MS is reached. The ions
tend not to be pumped away as much as the uncharged ions. At least a small
number are retained near the axis and can be analysed by mass-scanning the
quadrupole and detected. A multiplier detector is typically needed as the
number of ions is small.
The process of ionisation near atmospheric pressure by discharge is very
different from electron beam ionisation at low pressure. After initial
excitation or ionisation, which is primarily of the bulk majority component,
multiple collisions occur between molecules and molecules with low
ionisation potentials end up charged, whilst those (such as the inert gases,
nitrogen, hydrogen) with high ionisation potentials tend to end up neutral.
In this way the ionisation is highly selective for impurities and this is the key
to the extreme sensitivity of the method.
There are problems with APIMS, however. Like any MS, it cannot
distinguish from CO by mass analysis and at impurity levels above 100
ppb the ionisation mechanism becomes highly nonlinear in inpurity
concentration, making the method unusable except with ultrapure gases. A
Figure 2.42 Atmospheric pressure ionisation mass spectrometer.
sophistication of the simple APIMS arrangement provided by one company
(VG) offers a solution to the first problem. They provide a low-pressure gas
cell, which is actually another quadrupole mass filter (Q2) but is set to pass a
wide band of mass/charge ratio after the initial quadrupole (Q1), providing a
further analyser quadrupole (Q3) after the gas cell to analyse the resultant
ion stream. A mixture of and ions filtered out by Q1 collide with gas
molecules (such as Ar) in Q2 and produce and ions which are
analysed in Q3. Despite losses in the extra quadrupole stages, performance
of about 1 ppb CO in is claimed.
For calibration purposes APIMS needs a high-quality ultraclean dilution
rig. Calibration gases at ppb levels are difficult or impossible so higher level
contaminant gas samples are further diluted a factor of 1000 on a rig built
into the APIMS to provide suitable low levels for testing, using a built-in
purifier and bulk gas supply. The dilution rig can also be used to solve the
second problem above, that of the nonlinearity above 100 ppb. To measure
contaminants above the 100 ppb level, the sample gas is diluted with
ultrapure gas from the purifier by a further factor of 10 or more.
Time-of-flight (TOF mass spectrometry has been neglected in recent
years, reflecting its difficulties compared to quadrupole MS. However, it is
being revisited by a number of firms and products offering up to 350 atomic
mass unit analysis at low levels are now offered, including an API TOF
instrument offering 20 ppt detection limits.
The TOF principle is deceptively simple. If a brief, less than a
microsecond, bunch of ions is accelerated out of an ionisation chamber into
a vacuum with energy E, say 100 eV, then they will drift in an electric-field-
free region towards the collector at a speed v which varies as the square root
of their mass to charge ratio q/m
GAS TECHNOLOGY 181
In fact, the drift velocity is fast; for an ion, for example, it is
25 000 m giving a time of flight of only 5 or so in a typical flight tube.
The measurement must be registered on an oscilloscope for qualitative
assessment and then logged electronically with high-speed data handling
hardware, the mass spectrum being repeated hundreds of times per second.
High speed voltage pulses on the electrodes control the bunch formation and
sweeping out of unwanted ions and ion energies of up to a few hundred volts
are used.
2.3.7 Other analytical principles
Electrochemical cells. Electrochemical cells are some of the simplest gas
analysers and yet can be made to respond selectively to many common
gases, often with very high sensitivity. As mentioned in section 2.3.3, the
simplest electrochemical gas analyser is just two platinum electrodes dipping
into a cell of distilled water. If a basically inert gas containing is then
passed through the cell, the conductivity measured is proportional to the
concentration of in the gas. If the concentration of subsequently
falls, then is washed out of the water by the inert gas flow. Although no
longer used in practical instruments, this principle shows just how simple
electrochemical cells can be. The lead/alkali hydroxide/silver trace oxygen
cell is barely more complex than this but it is still very much actively used in
industry and is a model for many other similar electrochemical cells.
Designed for detecting trace oxygen in inert or reducing gases, the
lead/silver cell is capable of detecting sub-ppm levels of oxygen using the
following reactions.
At silver electrode, wetted by both gas and hydroxide
At lead electrode, submerged in hydroxide
Hence every molecule of oxygen which dissolves in the cell near the silver
electrode yields one electron, giving a very high sensitivity. Perhaps
surprisingly, there are few interfering effects to give rise to false readings.
Acidic gases will eventually neutralise the alkali solution and highly reactive
oxidant gases such as chlorine might also give trouble but these are relatively
uncommon contaminants, and can also be scrubbed out easily. Figure 2.43
shows how a simple electrochemical cell is constructed.
Slightly more sophisticated versions of this cell, using cadmium for the
submerged electrode and a third inert electrode, can achieve sensitivities of
a few ppb in nitrogen. To achieve this kind of performance, however,
these trace cells need sophisticated features such as pure nitrogen sparging
Figure 2.43 electrochemical cell.
of the reagents, temperature control and very careful sealing. Similar
electrode reactions using different metals and reagent solutions can be
devised for many other gases (section 2.5.2).
Zirconia oxygen analysers. Although the vast majority of electrochemical
cells involve aqueous electrolytes at room temperature, it is possible to run
an electrochemical cell with a solid electrolyte. If the solid electrolyte is hot,
allowing solid diffusion of ions to take place easily and rendering it
reasonably conductive, then workable analysers can result. The most
common analyser of this type is based on zirconium dioxide ceramic as the
active solid electrolyte. It has the feature that it measures not oxygen but
effective oxygen content. Most reducing gases, such as methane or CO,
will, if mixed with the gas being analysed, subtract from the oxygen signal as
if they were combined stoichiometrically prior to measuring on the
ceramic membrane. This is a useful feature where the analyser is being used
to indicated the oxidation potential of a furnace atmosphere, for example,
or where it is being used to adjust the fuel-air ratio of a combustion reaction
in a burner or engine.
The solid electrolyte is a thin plate of yttrium-doped zirconia ceramic
heated to 400C. Two porous precious metal electrodes are applied to the
ceramic membrane, with test gas and reference gas applied to opposite sides
of the structure. The arrangement constitutes a concentration cell. A
voltage V develops according to the Nernst equation, depending on the
logarithm of the ratio of the oxygen concentrations applied
where R is the gas constant, T is the absolute temperature and F is the
Faraday constant. For example, with atmospheric oxygen as reference and
100% applied, the cell will give 10 mV. A high impedance voltmeter is
connected to the electrodes and typically calibrated as concentration of
oxygen, assuming air is applied as the reference gas.
GAS TECHNOLOGY 183
Moisture measurement. The measurement of trace levels of moisture in
gases is both important and difficult. Much ingenuity has therefore been
expended on moisture measurement and there are dozens of different
designs. However, moisture measurements are still slow and inaccurate.
First, it is important to recognise that moisture is difficult to measure in
multigas analysers such as MS and GC. Moisture is also difficult to sample;
high levels of moisture will condense unless heated sample lines are used.
Before a steady reading of moisture is achieved, equilibrium between the
gas being measured and the pipeline must be reached but this is slow to
achieve. In long pipelines, even with no dead legs and high flow rate,
equilibrium is slow to achieve. Measurements of low levels of moisture also
ideally use heated lines to avoid errors caused by this invisible adsorbtion
along the walls of pipes.
In cryogenic liquids moisture may exist as tiny particles which are
stratified in the tanks. Like acetylene or higher hydrocarbons, water is
soluble only at low levels in cryogens. Depending on when samples are
taken, and where from within the tank, different numbers of these tiny
particles will be seen and different moisture levels recorded. Finally, in
cylinder gases, the moisture level increases as the pressure falls to low
values, i.e. as the saturated vapour pressure of the moisture on the cylinder
walls becomes a more appreciable fraction of the cylinder pressure.
The two most popular types of moisture instrument are oxide capacitance
and electrolysis. The capacitance instrument is of most general applicability,
measuring from a few per cent moisture down to ppm levels. It relies on the
fact that a specially prepared microporous oxide, typically aluminium oxide,
layer reversibly absorbs a small amount of moisture in a repeatable way,
with the amount absorbed being a monotonic function of the ambient gas
moisture level. A typical sensor has the layer grown on a conducting
support substrate, which is the back electrode. The other electrode is then a
porous conductor such as a thin gold layer. The electrical capacitance of the
assembly is linear in the amount of water absorbed and is easily measured
electronically. An alternative construction employs an layer on single
crystal silicon (Figure 2.44). The silicon can be conveniently equipped with
an integral heater, which enables desorption of water by heating prior to
measurement. This decreases the dry-down time and makes the silicon
sensor a more useful field instrument.
Electrolysis instruments are suitable for low levels (10 ppm and less) of
moisture. They consist of a closely spaced bifilar winding of inert metal on an
inert rod. The rod is coated with phosphoric acid. The acid dessicates at the
operational moisture level to phosphorus oxide, which is an insulator. If a
molecule of approaches, it is absorbed on the and releases two
ions, a hydroxonium and a phosphate ion, which are then discharged on the
inert metal to hydrogen and oxygen. The current flowing is thus pro-
portional to the amount of water absorbed by the assembly. The principle
Figure 2.44 moisture analyser.
was first developed to a usable instrument for measurement of moisture
levels in gas in nuclear reactors.
In repetitive processes it is possible to measure moisture changes which
are much too rapid to follow with moisture analysers by using specialised
sampling techniques. One set of work (Carr, 1994) measured moisture in a
rapid thermal processor (RTP), which is a small furnace for a single silicon
wafer used in the semiconductor industry. An RTP normally takes in a room
temperature silicon wafer and processes it at up to 1000C in a cycle lasting
just a few minutes. Moisture comes in with the wafer, from the air entrained
with the wafer and when the door opens, and can cause process problems.
To measure it, the IMEC team maintained a slow moisture probe in
equilibrium with a calibrated moisture-containing stream. The level of
moisture in the calibration stream was adjusted until a sample switched in for
a few seconds from the RTP did not disturb its reading. The technique is
analogous to a digital sampling oscilloscope in effect. Presumably another
way of obtaining readings from a repetitive process is to arrange for many
successive samples from a part of the cycle to be admitted, along with known
purity purge gas, measuring the average reading; in effect a stroboscopic
technique.
The so-called quartz microbalance can also be used for measuring
moisture. In this two quartz crystals are employed. Via their piezoelectric
effect, quartz crystals can be set vibrating at a very high constant frequency
depending only on the precise size of the crystal. Two crystals vibrating at
the same frequency are used such that if a small mass, even a microgramme
or so, is added to the surface of one, its frequency will fall slightly relative to
the reference crystal. A porous coating on the crystal means that even with
parts per million or less of moisture in the gas passing over the crystal some
moisture will be reversibly absorbed and give rise to a frequency difference.
The frequency shift, in the crystal resonant frequency, f, seen is given
approximately by
GAS TECHNOLOGY 185
where is the mass added over area A, is the density of quartz and k is a
frequency constant (0.17 MHz for AT-cut quartz). A 2 MHz crystal,
for example, might have a sensitivity of 200 Hz By using two identical
crystals and flowing reference and sample gas over the crystals (and
swapping over these roles every minute or two if necessary), most drift and
imbalance effects can be eliminated and the result is a very sensitive
moisture analyser which is very inert to chemical attack from reactive
components in the gas mixture sampled.
Analysis of scrubber solution. Although not strictly a gas analysis tech-
nique, the basis of many laboratory analyses for gases is still wet chemistry.
It is often convenient to convert a gas sample into a conveniently handled
liquid by bubbling it through a small scrubber or bubbler. Correctly chosen,
a scrubber solution may absorb only impurities and not the majority
components, providing a convenient concentration effect. No one would
measure with limewater in industry today. However, a search for, say,
trace metals in liquid would still most effectively be carried out by
absorption in acid followed by wet chemical reactions or ion chromato-
graphic methods.
Measuring silane purity. As well as monitoring scrubbing solutions, some
gases can be reacted with solids which can be analysed to give gaseous
purity. One example of where this kind of technique yields an effective
sensitivity which is higher than any direct analytical technique is in the
estimation of silane and chlorosilane purities. These gases are frequently
used (section 3.5.6) in the semiconductor industry to lay down epitaxial
crystalline layers on top of semiconductor silicon. A measure of the purity of
the achieved deposited layer is, in effect, a measure of the gas purity: if the
layer has no impurities, its electrical conductivity will be very low, caused
only by thermally excited holes and electrons. However, donor and acceptor
impurities, such as gallium or phosphorus, donate electrons or holes, which
conduct electricity, to the silicon matrix. This turns out to be a very sensitive
analytical method. Only 1 ppb of gallium, for example, would give rise to
conductivity 3000 times that of intrinsic (theoretically completely pure)
silicon. Sometimes, (chloro)silane gases are actually specified by the
resistivity achievable, rather than by actual gas purity. With a baseline of
something like 400 000 for intrinsic silicon, a gas with 1 ppb donor
impurities might in the worst case be a gas describable as 130
The purer the grade of silane, the higher the resisitivity. Typical values vary
from 1000 for an ordinary grade to 10 000 and more for an
ultrapure grade.
The extremely high sensitivity of this technique raises the question of
whether it can be adapted to measure other gases. In a sense, tin oxide
semiconductor bead flammable gas detectors are using it because they
actually work because of the production of carriers by a redox reaction in the
bead. Analysers using a negative effect are also conceivable in which a thin
layer of doped semiconductor changes its resistance after the gas reacts with
and removes donors or acceptors.
Chemiluminescent (light-emitting reaction) analysers. There are a number
of gas analysers which rely on the emission of light by the gas being sensed.
White phosphorus suitably prepared glows as it oxidises when oxygen is
present. The light emitted at low trace levels is proportional to the amount of
present and can be used down to sub-ppm levels for the detection of trace
The arrangement for a gas analyser is simply to feed gas into a blacked,
tortuous pipe (to eliminate stray light) inside which there is a disc of white
phosphorus facing a photomultiplier tube a few millimetres away. The
phototube detects the few visible photons emitted by the reaction and boosts
the signal to a level of reasonable current.
Many analysers also rely on chemiluminscence, in this case the light
emitted when reacts with ozone, also achieving a high level of
sensitivity.
Mercury reduction analysers. The modus operandi of these is to pass the
sample gas mixture through a tube containing mercuric oxide. At a few
hundred C, any reducing gas present will cause the release of minute
amounts of mercury vapour. Mercury vapour can be detected very
sensitively (in ppb or even ppt) by its very high optical absorption, which is
especially high for optical emissions from a mercury lamp. Although rarely
used on their own, they are found in specialised gas chromatographs where
they can monitor for and CO while ignoring inert major components such
as or Ar. The simpler FID detector could not detect would require a
methaniser to pick up CO and might be swamped by with a methaniser.
Magnetic oxygen analysers. These rely on the extraordinarily high mag-
netic forces which oxygen gas shows in a magnetic field, the paramagnetic
susceptibility of oxygen, which is far higher than any other gas. Oxygen
molecules behave as small dipoles, i.e. in a magnetic field they have a force
exerted on them which is proportional to the field gradient and in the
direction of increasing magnetic field.
Some instruments simply place a tiny lightweight hollow dumbell (filled
with a gas other than oxygen, or under vacuum) in the gas stream, with one
half of the dumbell in a strong magnetic field. In the event that the sample
stream contains oxygen, a force is exerted on the dumbell, since the sample
stream is then attracted into the magnetic flux, whilst the dumbell, being
GAS TECHNOLOGY 187
Figure 2.45 Magnetic wind analyser.
magnetically inert, is not and is therefore repelled. The force of repulsion is
measured by feeding back from an optical position sensor a twisting force via
moving coils as used in electrical meter movements such that the dumbell
does not move. The force, F, exerted is approximately given by
where X is the paramagnetic susceptibility, A is the area of the dumbell in
the non-uniform fringe of the magnetic field and B is the magnetic field.
Some feeling of how difficult these instruments are to construct can be
gained by noting that in a 0.1 Tesla field (the most readily achievable with a
modest permanent magnet), a 20% sample will only exert a net force of
order on a dumbell of 5 mm diameter.
A simpler type of instrument (known picturescuely as magnetic wind
oxygen analysers) detects flow through a heated pipe in a solenoid field, and
depends on the fact that the paramagnetism of oxygen is temperature
dependent (Figure 2.45). Again the wind force measured is very small,
corresponding to a differential pressure of the order of microbars. The
results from this form of instrument are affected by temperature, gas
pressure and carrier gas type as well as oxygen concentration, so careful
calibration and correction algorithms need to be applied for accurate work.
Other optical absorbtion analysers. In the section on IR analysis above it
was noted that some gases at least absorb electromagnetic radiation of other
frequencies. The greenish hue given to the air by that is due to absorbtion
in the optical wavebands, for example, can be detected electronically. There
are microwave rotation absorption bands, well-documented by research on
high resolution military radar. For example, oxygen has absorption lines at
5000 However, special techniques would be needed to effect detection:
RF circuits at 60 GHz are relatively difficult, secondly, the maximum
absorption is only of the order of 10 dB per kilometre, or 0.01 dB (0.23%
absorption) in a practical size, say 1 m, resonator. Experimental oxygen
analysers have been used using the ultraviolet (UV) absorption of oxygen.
While is a non-absorber, has a high absorption starting at 190 nm.
Claimed advantages are simplicity, high speed of response, useful in medical
applications, and high selectivity to oxygen, although no commercial
analysers on this principle are available currently.
2.3.8 Monitoring air for gases
In addition to the commercial requirements for analysis, the gas industry
and gas users have a responsibility to monitor the air for emissions
hazardous both to the health and safety of their staff, and also for emissions
ultimately harmful to the wider environment.
22
Oxygen depletion monitoring. Oxygen depletion monitoring is employed
to prevent exposure of personnel to air with too little oxygen. Depleted air is
relatively common (section 2.5.2) and very dangerous. Just a few breaths of
depleted air, a matter of a fraction of a minute, is dangerous so monitors
should ideally be mounted near the mouth of persons to be protected and
should have a response time of a few seconds.
These requirements are difficult to satisfy. Only electrochemical cells
normally meet the portability requirements of small size and power
requirements, whilst only instruments such as the paramagnetic oxygen
detector have really quick response times.
23
However, the latest generation
of electrochemical sensor is faster than ever and meets the requirements of
safety in many cases.
Flammable monitoring. Safety requirements frequently lead to a demand
for rapid response instruments which respond to all inflammable gases and
vapours. Ideally such universal flammable gas monitors should show a
response proportional to the lower explosive limit.
The heated metal oxide semiconductor bead detectors, such as those
pioneered by Taguchi in Japan, are now ubiquitous. They function by
registering a change in electrical conductivity. An oxide, often tin oxide
based, bead is heated to 300500C. On exposure to a flammable gas in air
the electrical conductivity increases. The mechanism would appear to be a
catalytic oxidation resulting in charge carriers being injected into the oxide.
Although inexpensive and useful in alarm circuits, their quantitative
22
Science and industry used small animals to monitor air quality unt i l quite recently. British
mines had canaries, for example, while Priestley (Gibbs, 1965) describes how the number of
mice used in experiments could be minimised by keeping hold of their tails and withdrawing
them when they appear stressed. The practice is not. however, really acceptable today.
23
As noted above, an ultrasound device, which can be made portable and rapid in response, is a
possible option for some situations.
GAS TECHNOLOGY 189
accuracy is, despite many years of development, dubious. First, the output
from such detectors is not particularly stable. Second, water vapour often
gives some spurious response, the response to non-hydrocarbons such as CO
is often poor and poisoning reactions are common. Particular poisoning
reactions are frequently seen with reactive silicon-bearing materials such as
silicone aerosols (commonly used as polishes) and with chlorofluorocarbons
(still used in some specialist aerosol cans and in cleaning).
Toxic monitoring. Much toxic gas monitoring is still performed by
chemical tape detectors. For nearly all gases a reaction can be devised
whereby the contaminant produces a colour change on a piece of porous
paper impregnated with a reactive solid. A very simple example is the
monitoring of the highly poisonous hydrogen sulphide which, despite
apparently being easily smelt at low concentrations, can be dangerous
A roll of tape is unwound slowly from a reel across an active aperture where
air either diffuses or is pumped. A photodiode and lamp then checks the
colour of the exposed tape and rings an alarm if a critical level of coloration is
exceeded. The exposed tape is then wound on to a take-up spool where it
constitutes a reviewable record of the toxic gas concentration.
Simple electrochemical cells are also excellent in this job. Rather than
pumping the air sample across the electrolyte as in analyser practice,
however, typical cells allow the sample to diffuse into the cell. The cell is also
typically sealed from moisture loss by means of a semi-permeable polymer
membrane. Polythene and polytetrafluoroethane, for example, are ex-
cellent moisture barriers but allow many gases to pass readily. Many
different cells are available, covering many of the common requirements for
toxic monitoring. For example, one manufacturer lists cells available for the
following gases in air:
HC1, HF, These cells are all claimed to operate at sensitivities
of 1 ppm or less. Most are based on a simple two-electrode electrochemical
cell, choosing the electrode and electrolyte to selectively respond to the gas
of choice. A typical example is the CO cell. This has the reaction
at the sensing anode, a porous conductive type. The electrolyte, a strong
sulphuric acid solution, conducts the hydrogen ions away and ensures that
the is not absorbed but can diffuse away.
Although many of these cells cannot make accurate measurements and
are subject to many interfering effects, they excel in the role of air-
contamination alarms. Exact quantitation is not needed. Also, provided
that interfering gases do not interfere in a negative way, subtracting from the
true reading, and provided that false alarms are not produced by normal
concentrations of innocuous gases, interference is not important in this
application.
The principle of photoionisation by UV mercury lamp illumination has
also been used for toxic gas monitoring. Photoionisation detectors respond
to any contaminant in the air which has an ionisation potential below that of
the UV photon energy. With a photon energy of 10 eV, air gases and some
pollutants are not ionised but many important pollutants are, such as
and Photoionisation is perhaps best used as a GC detector for
detection of hazardous gases at trace levels in air, however, because the
addition of GC separation enables a more confident measurement. The fact
that the detector does not respond to air gases Ar, and
means that it can achieve a high sensitivity to these and can use filtered/
purified air as a carrier gas if necessary.
The remote sensing of pollutants in air by laser Raman spectroscopy was
mentioned in section 2.3.5 but this sort of instrumentation is not common in
everyday industrial use because of its cost and complexity.
Mass spectrometer multipoint systems. These rely on a central mass
spectrometer analyser connected by a multiway valve to a series of sample
pipes. Typically a central display unit plots the samples versus time for each
sample point. In addition, an electronic signal representing each measure-
ment made is typically returned to a display next to the sample point.
Great care must be taken to make sure that the toxic gases monitored are
not absorbed by the pipework system and to make sure that the pipework
does not induce undue delays. Because of these considerations, the pipes are
continuously pumped to maintain flow and improve the response time of the
instrument.
Heated sample lines are also popular because they avoid problems with
moisture condensation, for example by continuous heating of the sample
tubing. A convenient way of achieving this is to pass a current down a
stainless steel tube such that it acts as a resistor. With some thermal
insulation around it, this can be a very convenient system. Even more
convenient is a coaxial assembly with a narrow-bore stainless steel sampling
tube down the middle, which is insulated and a copper braid and outer
covering used. The heating current from the low-voltage transformer travels
up the braid with little loss and the stainless steel is heated in the current
return path, the sampling end of the tubing being soldered to the braid. In
this way, the heated sample line can be used just like a simple tube, can be
flexible to some extent and can be installed more easily.
Occasionally, the central mass spectrometer will use a membrane
separator. A dimethyl silicone rubber membrane will effectively remove
much of the air from the sample stream, whilst absorbing and transmitting
many toxics to the MS source region (see remarks on the PETRA
GAS TECHNOLOGY 191
instrument in section 2.1.7). Although quantitative measurement is then
more difficult, the concentration factor makes the instrument much more
sensitive and this may well be the decisive factor, rather than accuracy, in a
toxic monitoring application.
The human sense of smell. The sense of smell as a multigas analyser is often
overlooked. The nose is not a quantitative analyser, tires easily (Moncrieff,
1967), does not detect asphyxiant and many more dangerous gas mixtures,
and is thus mainly useful for qualitative monitoring and toxic gas warning. It
is, however, enormously important. Were analyses of a large enough
number of industrial accidents involving gases carried out, they would
almost certainly show that more lives are saved each year by people using
their noses than are saved by the most sophisticated monitoring systems;
which is not to say that monitoring systems are not also essential. There are
also many industrial processes where smell is an important process or quality
control input. Efforts should be made both in the design and operation of
plant to use the sense of the smell as a safety feature. For example the design
of systems should avoid, where other considerations allow, the routine
presence of masking odours that might prevent a leak or other malfunction
being detected via smell. Similarly, operators of plant should never ignore
unexplained smells. (A brief introduction to the sense of smell is given in
Appendix B.)
2.4 Gas handling and storage
The generation of large amounts of electrical power was already possible by
the mid-nineteenth century. However, the widespread use of electricity was
much delayed because the means to deliver that power were not sufficiently
developed. Even the comparatively simple requirement for a well-insulated
cable was not easily met at that date, and the many familiar ancillary devices
needed (transformers, fuses, switches) were not available in suitable form.
Transport of gases from the point of production to the point of use is no less
important. Gases must be delivered to the point of use at the correct
temperature, pressure, flow rate and chemical composition, and a host of
analogous ancillary devices are needed, e.g. pipes, regulators, bursting discs
and valves.
Fortunately, gases can be efficiently stored, contrasting with the diffi-
culties that attach to electricity storage. Storage has meant that pipelines
from the point of production to the point of use, except for coal-gas and,
today, for methane fuel, are not generally needed. Originally gases were
simply stored at low pressure in gasometers, which are simply water-sealed
piston and cylinder arrangements. Today, efficient storage relies on storing
cryogenic temperature liquid gases at low pressures or storing gases in high
pressure cylinders. Industrial gas for a typical application is therefore
produced, delivered either in cryogenic form or in high pressure cylinders by
road, then piped a short distance, via pressure and flow controls and perhaps
via filtration and purification, to the point of use. Gas pipelines, for all that
they are usually short, however, still require very careful design.
The materials used with industrial gases vary from the unsophisticated
black iron pipe to high technology oils such as the perfluoropolyethers. The
choice of materials for pipes, seals, lubricants and other gas-wetted
components depends on the gas and pressure to be contained. Gas suppliers
generally give tables of materials recommended for use with different gases
and these should be consulted during the design of a system. These tables
can be supplemented with information from suppliers of special materials,
such as elastomer seals, who will often have carried out extensive testing for
suitability.
2.4.1 Pipelines and pipeline components for gases
Pipelines. The extensive use of steel black iron pipe for long-distance
pipelines for (relatively) inert gases in bulk depends on its strength and very
low cost. For dry gases inside, and with paint and perhaps electrical
protection on the outside, it is a highly satisfactory material.
Choice of materials is always a combination of technical suitability, cost
and convenience in application. The extensive use of copper tubing for small
bore installations of all description for many different gases is largely due to
the latter factor. Copper tube is by far the easiest material to handle and
install for lower pressure work, has a fairly low cost and is chemically inert to
nearly all gases. Stainless steel tubing has been slowly growing in popularity
for many years. Soft-annealed stainless steel tubing can be handled nearly as
easily as copper, while it will retain an attractive external appearance, keep
very clean inside over time, and can be used with many gases. Its use is
routine in semiconductor gas installations, being valued for its low
particulate and corrosion resisting properties, although even semiconductor
installations historically used copper pipe without problems for inert gases.
Gas normally contacts (wets) only the inside of a pipe or vessel. It is
therefore not necessary to construct pipes or vessels out of highly inert bulk
materials; a coating or lining, if sufficiently thick and contiguous, will often
suffice. Although the relatively small sizes of pipes used in industrial gases
usually preclude any kind of liquid-based coating process, gaseous passi-
vation processes are used, for example fluoride passivation with a weak
fluorine in argon mixture.
Metal pipes are made in many different ways: extrusion, rolling followed
by welding and hot or cold drawing down of a short fat pipe are those seen
most commonly. For the more easily melted softer metals, such as
GAS TECHNOLOGY 193
aluminium or copper, extrusion is popular, whilst welding is the cheapest
method for thin-walled steel tube. The very highest quality pipes are today
made by the drawing down of a short fat pipe called a bloom. This is
produced by forcing a red-hot billet against a mandrel, which pierces the
billet. Cold drawing can yield very high-quality, strong pipe.
The methane fuel gas industry has recently taken to installing small and
large diameter (from 20 mm up to 500 mm diameter) simple unreinforced
polymer pipes for underground transmission of gas. These pipes are made of
polyethylene (PE) and are tough, inexpensive, and completely rustproof.
Underground, they are estimated to have the potential to last for a minimum
of 50 years. Part of their attraction is that they can be heat welded in situ. The
heat welds in PE are easily as strong as the original material and are much
less likely to leak than, for example, jointed iron pipes. Welds are achieved
by cutting the pipe ends off square, heating them by contact with an
electrical hotplate and then ramming the two softened pipe ends together
with a hydraulic rig to form a butt joint.
The PE material has a usable strength of about 10 MN giving them
potential pressure ratings of the order of 20 bar in pipes with walls which are
not too thick. In fact, however, methane distribution is low pressure (100
mbarg) to domestic users and only 2 barg or so to normal industrial users,
allowing a generous safety margin.
So far, few industrial gas users have taken advantage of the possibilities of
pipes made from these materials. There are concerns about moisture,
hydrocarbons and other contaminants, for example, getting into pure gas
streams, whilst the use of plastic pipes with oxygen has obvious safety
problems and the pressure ratings developed so far have been lower than are
generally used in industrial gases. However, they are now a tried and tested
system, offering considerable advantages for installations with very long
runs of pipe.
Flexible pipes. Standard rubber hoses, consisting of fabric-reinforced
synthetic or natural rubber, are normally made with lower pressure ratings,
e.g. 20 bar or less. Although rubber hose with some higher pressure ratings
is available, these materials are not particularly popular with industrial gases
because they are heavy if they are strong enough and subject to degradation
from UV or oxidation.
For flexible connections in systems, bellows hoses, typically made out of
stainless steel, are often employed. Care should be taken with bellows that
they are not cracked. If they are subjected to a sharp bend and then bent
back again they can fail; in applications where repeated flexure is needed,
the amount of flexure needs to be strictly controlled within the limits which
the supplier gives. In addition, for lower pressures, straightforward
unreinforced tube in nylon or another strong polymer can be used. Simple
polymer tubing is certainly very convenient, even if it is semi-rigid, for
example for pneumatic control connections, and there are several inex-
pensive instant connect systems which allow very simple connections for
pneumatics. However, the rapid loss of mechanical qualities with tempera-
ture in polymers should be borne in mind; nylon becomes surprisingly soft
even in hot water. Ratings of up to 10 bar are readily available, however, in
unreinforced tubing.
For high-pressure flexible connections, a bellows can be armoured with a
multilayer braded stainless steel overcoat, giving pressure ratings up to
200 bar or so. More cheaply, a pigtail or coil of a smallish bore pipe is
commonly used and, if a reasonably long length of pipe is coiled, this is
surprisingly flexible. However, for many applications genuine flexibility is
needed and this is where armoured polymer hose is mostly used. This is
supplied with special crimped end-fittings by the manufacturers, and
normally comprises two or three layers. The inner tube, containing the gas,
is made from a polymer such as nylon, which is both strong and also a good
gas barrier. Teflon is often used because of its chemical inertness and
therefore wide applicability with nearly all gases. The middle layer is a
strength layer, typically a fibre-reinforced rubbery polymer, or it may be
braded stainless steel. The outer jacket, if used, protects from abrasion, UV
degradation and oil or solvent attack. A neoprene or polyurethane material
is typical. Polymer hoses can offer pressure ratings up to 300 or 400 bar.
Patented quick-connect systems are available for use with any pipework
but obviously are particularly aimed at use with hoses. A quick-connect aims
to provide a gas connection within a second or two, typically by simply
pushing the pipe fittings together, perhaps with a twist or disengagement
ring to pull to lock them in place, and also has provision for automatically
sealing off the ends of either or both of the pipes when the joint is
disassembled again.
Pipework joints. The most satisfactory pipework joints for many purposes
are those which braze or weld pipe sections together, provided the minimum
possibility of a leak developing subsequently and maintaining the pressure
rating of the pipe itself. The lead/tin soft soldering of copper pipes used in
low-pressure domestic gas installations is unacceptable in the much higher
pressure system pipelines used in industrial gases in the latter regard.
Instead, copper tube may be brazed, using similar joints to the more familiar
soldered types, but using the much stronger and higher melting brass-based
brazing alloys instead of solder for the filler metal.
Although brazing can be used on stainless steel pipes, these are generally
joined by tungsten-inert gas (TIG) orbital welding, using automatic
machines from companies such as ARC. Orbital welders, once correctly set
up, can perform butt welds on prepared pipes in a matter of seconds, with
l i t t l e in the way of fumes or fuss, and the joint is bright and almost
indistinguishable from the pipe itself from a distance. They clamp the two
GAS TECHNOLOGY 195
pipe ends tightly together, apply argon inerting then, under computer
control, a TIG tip arc is struck and the tip is then orbited around the pipe at a
precise speed, stopped as the arc returns to the starting point and the joint is
completed. A shield renders the wearing of goggles unnecessary and the
amount of heat used is surprisingly low; the welded joint can be handled in a
matter of seconds. Pipework can also be welded by using socket welding.
Fittings are arranged to have a socket to match the pipe used, the socket
supporting the joint while welding takes place. Automatic TIG welders can
also usually handle socket welds. Some fittings are arranged to be a
compromise between butt weld and socket weld, using a very short (1 or
2 mm) socket, which aids in providing a little extra material as a filler in the
completed weld but does not have the disadvantage of containing a crevice
(which could harbour contaminants) in the finished joint as there is in a deep
socket weld.
For joints in larger pipelines, it is clearly appropriate to use welded-on
flanges, which may then be bolted together with a flat composition sealing
washer or perhaps a more complex seal. Flanged joints are highly
inconvenient, however, except in the largest sizes, taking much time and
labour to assemble for the first time and for subsequent assembly/
disassembly, and must be tightened carefully and evenly to be effective. For
slightly smaller pipes which can be handled suitably, at modest pressures, it is
of course possible to cut tapered threads on the end of the pipe and joint them
using bushing or union nut couplings. For long pipelines made of inexpensive
black iron pipes, this method is often used. However, cutting threads on the
outside of a pipe reduces the pressure rating of the pipe considerably.
Conventional compression fittings of the type are very widely
used (there are many other brands with a variety of pressure ratings and
prices) (Figure 2.46). These are convenient and can be made highly
leak-tight. Their convenience lies in the fact that the pipes need no special
preparation other than cutting to length and no special equipment is needed.
Also, once the joints have been made, they can be easily demounted and then
reattached. The tubing must have a good outside surface finish, must be kept
free of scratches, and the sawn-off end must be carefully deburred, however.
The tightening of the fitting itself, when first tightened, attaches the olive
(the sealing and mechanical restraint ring) firmly on, giving a joint which will
meet very high pressure ratings to match comparatively thick walled pipes.
Standard 6mm pipes in stainless steel with such joints can be used up to500 bar
or so. Their modus operandi is to crimp the olive down onto the pipe by an
annular wedging action. Because the pipe material is selected to be softer than
the fitting material, the olive is crimped inwards onto the pipe, which is slightly
deformed inwards, retaining the olive. Care should be taken not to
overtighten, however; the joint will not be more leaktight, and the female
cone of the fitting can be splayed out, making it difficult to demount and reuse
on subsequent occasions.
Figure 2.46 Compression fi t t i ng used for joining pipes in the gas industry.
Fittings of this type are available from (2 mm) through to pipe
sizes. For the larger sizes pressure ratings are lower, although with rather
thick-walled pipe 200 bar ratings can be achieved. A strong arm and large
spanner is needed to swage the olives onto the pipes in the larger sizes and
electric or hydraulic machines are available to render this task easier. These
have the additional advantage that no strain need be placed on the fitting to
which the pipe wi l l be attached, other than the more modest torque required
to effect a gas seal.
With all these proprietary joints, the manufacturers tables of fittings and
fitting/pipework combinations should be consulted as to the usable pressure
rating and piping to match the fitting manufacturers requirements should be
obtained.
High pressure pipeline and storage vessel design and materials
Pressure vessels. A pressure vessel can be taken to mean a container for
gases of above a certain size (say 0.5 1 or so) and above a certain pressure
(perhaps as low as 1 barg or less). After many spectacular and usually fatal
accidents in the nineteenth century (e.g. Anon, 1993) with exploding
boilers, pressure vessels are now subject to a large number of regulations.
The regulatory codes are listed by government authorities across the world,
with many countries adopting the ASME codes of the USA (ASME VIII
part 1), the DIN codes of Germany or the BS (BS 5500 for fixed vessels, BS
5045 part 1 for seamless and part 2 for welded movable vessels) codes of the
UK. The codes do differ in some respects, but are broadly similar.
ASME and the other codes specify:
materials: a limited number of metals, mostly grades of steel, are accept-

able;
GAS TECHNOLOGY 197
safety factors;
structural design, calculation methods and record-keeping;
design details, such as welds and fittings;
inspection and testing;
corrosion and fatigue life;
conditions of use (pressure cycling, temperature, pressure, intended
contents and trace contaminants);
ancillary equipment, such as pressure relief valves.

The codes themselves should always be consulted before constructing a
vessel, although pressure vessel suppliers always work to one of the codes
and generally will work to any code on request. However, it is probably
worthwhile noting at least some of the details contained in the codes:
dished ends are produced by rolling of flat plate with a high power
hydraulic ball forming system: this does produce some thinning of the
plate used and codes specify what allowance must be made for this;
longitudinal joints in cylinders are under higher stress than circumferen-

tial joints and this must be accounted for;
vessels must show a safety factor of 4 versus maximum design pressure

before failure and must show a safety factor of 1.5 versus maximum design
before plastic flow (yield) (ASME);
where connections to pipework, manholes or other devices pierce the

vessel wall, formulas are given for putting in enough strengthening
material to compensate for the loss of wall material;
impact testing such as Charpy V notch at a low temperature may be

specified for materials that can become brittle.
Cracks and testing. It is very rare indeed for a pipeline or gas cylinder to fail
under ordinary conditions. Materials selection is one reason for this but an
additional reason is that materials, components and complete installations
are all tested.
All cylinders, many other components and some pipelines are tested by
filling with water and pressurising well beyond the design pressure. As noted
above, hydraulic testing is comparatively safe because if failure happens to
occur, the energy release is low compared to the explosive release that
follows failure with gas. Hydraulic testing, however, has its own problems.
Water or other fluid must be drained out and the vessel dried after testing.
This presents problems with some designs, which must be drainable with
water-compatible components. Furthermore the temperature of testing
must be carefully chosen. If it is low, then the design may fail due to brittle
failure but this may constitute a better proof test. If a higher temperature is
chosen, then failure is unlikely but some amount of distortion may occur if
the material yield stress is exceeded.
Another important means of testing is that of crack detection. Two simple
Figure 2.47 Crack detection with a magnetic sensor.
techniques for finding cracks which extend to the surface are sprinkling with
iron filings and the use of a fluorescent dye penetrant. The dye is sprayed on
to the surface to be tested and any excess wiped off. A crack retains some
dye which then bleeds out on to the surface and becomes visible.
Iron filings sprinkled on to a magnetic vessel or pipe which has some
magnetic flux imposed on it from an external magnet or coil will tend to
accumulate where there is a rapid change in flux. Where there is a crack,
even quite a small one hidden by subsequent painting, iron filings will tend to
stick since the crack causes a big change in magnetic flux.
More sophisticated than these, but still using essentially simple equip-
ment, are electric or magnetic methods (Figure 2.47). The latter amounts to
little more than an instrumental method to replace iron filings. A trans-
mitter which induces a d.c. or a.c. magnetic flux in the test piece is attached
by a clamp to the latter and a search coil shows where d.c. or a.c. magnetic
flux at the surface changes rapidly. An eddy current detector (similar to the
treasure-hunters metal detector) can be used as an a.c. magnetic
instrument. The eddy current detector uses a search coil to induce an a.c.
field in the material being examined and this (typically) changes the
self-inductance of the search coil and hence the a.c. generator frequency.
Any change in frequency signifies a problem. Cracks only 0.5 mm deep can
easily be detected but further sensitivity is typically inhibited by the surface
finish on the vessel examined. Electrical resistance methods are rarely used
but these can detect many flaws in pipelines or vessels in a very simple
fashion, if care is taken with contact resistance problems, for example by
GAS TECHNOLOGY 199
using multiple finger contacts and four-point (or more) resistance measure-
ment.
X-ray inspection. X-ray radiographs (sometimes actually made with a
radioactive source such as cobalt-60) are now a standard routine for welds,
at least for a sample of welds carried out. X-ray equipment is fast and special
X-ray tubes are available for different geometries of weld; for example
sources emitting a disc-shape fan of rays for irradiating circumferential
welds in pipes and vessels with one shot.
Gamma-ray sources such as cobalt-60 require longer exposure times and
more sensitive film than X-ray radiographs. However, they can provide
images of thicker pipes and can be very small and lightweight, enabling them
to be pushed up inside even quite small pipes (with precautions to avoid
radiation exposure to personnel). They can also be mounted on a small
motorised crawler or pig which can be manoeuvred hundreds of metres
along a pipeline.
Weld radiography is required in many codes for pipes and vessels carrying
toxic materials, and is good practice where any doubt whatever exists on the
quality achieved in welding. However, it is worth perhaps pointing out some
of its limitations:
the need to clear personnel away from a radiography site is awkward in

situations such as a construction site;
off-line processing of film is time-consuming;
image quality is often limited by fogging (due to scatter of X-rays), grain

(due to low achieved dose) and source size;
thicker samples need higher voltage X-ray tubes or cobalt-60.

Ultrasonic crack detection. Ultrasonic systems operate on the time-
honoured principles of military radar or sonar, sending out a short burst of
MHz ultrasound via a piezoelectric transducer, then listening for reflections
from it. With the speed of compression waves (the main transmission mode)
in steel 6000 m or so, an A scan of a 5 mm thick pipe wall will be only
1 long.
24
Commercial systems are as little as $10 000 or less to buy,
although a computer-based system with two-dimensional plotting and signal
processing might be $50k100 000. They are particularly suitable for
material without welds and, of course, provide an accurate readout of metal
thickness.
A particularly useful kind of ultrasonic transducer is that which can launch
a wave at an oblique angle to the surface. This can then be placed on the
smooth walls outside of any metal-working zone and used, for example, to
24
This sounds difficult but in fact the electronics is fairly easy with todays electronics; the
author made a special purpose system of this kind to work with a laboratory oscilloscope in a
few days.
Figure 2.48 Cracks in A-scan detected by acoustic reflection.
examine a neighbouring weld zone. This kind of transducer is used on gas
cylinders.
The principle difficulties with ultrasonics often revolve around the skill of
the operator in interpreting the images and oscillograms obtained, and in
maintaining good sound-conducting contact with the test piece. Unfortu-
nately, the thinnest imaginable air film between the transducer and the test
piece completely reflects the transmitted signal and prevents reception of
the ultrasound signal; this is why the system detects small cracks so easily. A
thin film of oil or gel is usually necessary to ensure an adequate contact with
the substrate being examined. Also, there are often a surprising number of
spurious reflections echoing around a pipe or cylinder to cause confusion to
the unwary, so interpretation is quite skilful, although multiple tests on
similar objects like cylinders do allow a fingerprinting approach to be
taken. Computer-based testing of gas cylinders can now be carried out by
relatively unskilled operators, using a computer to interprete the wave-
forms, knowing the types of reflections obtained. Computer systems also
allow the display of a two-dimensional map of reflections, giving a rather
better idea of where defects are than simply chalking on the cylinder where
the A-scan showed a result.
25
Figure 2.48 shows the principle of A-scan
acoustic reflection for crack detection.
Acoustic emission (AE) is one of a number of more unusual methods
which are also occasionally used for tests of cylinders. A crack does not
25
Those who have seen ultrasound in action for antenatal care in hospitals will no doubt have
been duly impressed by the pattern-recognition skills of hospital ultrasound technicians. They
can confidently comprehend the apparently meaningless blobby image and assess the
development of a human foetus with surprising accuracy.
GAS TECHNOLOGY 201
lengthen smoothly but progresses by small increments related to the grain
structure of the material. The progress of a crack as pressure is increased on
a cylinder can actually be detected by picking up acoustic impulses which
occur as the crack elongates. Furthermore, by timing the arrival of the
impulse at each end of the cylinder, the position of the expanding crack can
be detected. It is only very recently, however, that acoustic emission has
been sufficiently sensitive to be useful in the gases industry in small
equipment. AE equipment is slowly becoming accepted and may eventually
be accepted as a standard test method for cylinders and pressure vessels,
along with hydraulic testing.
The size of crack that is permissible depends on the alloy used, as well as
on the pressure of the service required. The alloys used for gas cylinders are
chosen not just for their strength but also for a high work of fracture per unit
crack area, W. With simple assumptions, the maximum size of crack,
can be estimated approximately from
where E is Youngs modulus and s is the average tensile strength. Beyond
this critical crack size, according to the simple theory, the energy generated
by further extension of the crack exceeds that needed to create more crack
area so the material fails catastrophically. In a typical example, the
maximum size of crack in typical mild steel sheet is an appreciable fraction of
a metre by this formula clearly well beyond the size at which a leak would
be clear but the formula is still useful to remind us that the smaller the work
of fracture, the smaller the critical crack size. Thus, because the work of
fracture for brittle metals such as the high strength, high carbon steels is just
a few thousand J they are unsuitable for cylinder manufacture.
Another potential problem is fatigue cracking. This is typically a problem
with aluminium and has been extensively researched because of the
extensive use of aluminium in aircraft. Given the fairly uncontrolled
conditions of use of the hundreds of millions of gas cylinders in use
worldwide, extreme caution has been exercised in the use of Al alloy
cylinders. In fact, this conservatism has led to Al cylinders having very thick
walls, and being about the same weight as a steel vessel of the same volume
except at the smallest sizes (section 2.4.7). Nevertheless, some difficulties
have been seen in aluminium cylinders, typically due to problems not related
to the basic cylinder containment problem. Corrosion due to alkali
exposure, fairly harmless to steel cylinders, and neck-cracking caused by
excessive stress induced by the valve fitment are two of these.
The embrittlement of metals caused by hydrogen is another potential
problem. Steel cylinders for hydrogen are specially heat-treated to minimise
this effect and are inspected especially frequently. Hydrogen pipelines, at
least in smaller sizes, frequently avoid this problem because they are made in
thick-walled cupro-nickel alloy (tungam) tubing.
Pipework design. Industrial gases such as nitrogen are frequently piped
around sites to multiple user points as much as other commodities such as
hot water or electricity. Indeed, some of the same principles apply to and
other bulk gas installations, for example:
pipe sizing should be such that undue pressure drops are not seen with all
equipment operating (also, pipe sizing should allow for additional
requirements such as the gas velocity speed limit for oxygen,
when multiple drop points are needed, for example, a ring main offers
the possibility of smaller bore pipe than a similar single-ended pipe;
on-line (pressurised) connection/disconnection facilities, as provided by

outlets with isolating valves or, better, with self-sealing quick-connects,
are an advantage, enabling a new equipment to be connected without
interruption of other equipment.
Pipework must take into account often conflicting demands:
convenience in use: pipework should ideally be easily accessible for

maintenance purposes, for example;
regulations on pipe locations (fuel gases should not be run near to

oxidants, for example) and labelling;
ease of fabrication: anyone who has tried to weld or solder a pipe in an

underground or underfloor run will attest to the difficulties of some
pipework layouts;
pressure losses: an excessive number of sharp elbow bends, for example,

or the use of too many high loss seat valves, for example, can give rise to
high pressure drop over a complex system;
in liquefied gas installations, in addition perhaps to cryogenic problems,

the ordinary features of liquid pipework, such as water hammer
26
or
cavitation effects, must be avoided.
Where multiple similar units such as filters or heat exchangers are
connected together in parallel with pipework, it is possible for an undesir-
able maldistribution of flow to occur and the pipework designer should be
aware of this. For example, in heat exchangers it is good practice to arrange
parallel units in Z-manifolds rather than U-manifolds in order to avoid
maldistribution.
26
Water hammer occurs typically when a valve on the end of a long pipeline running at speed is
closed too quickly. A potentially damaging high pressure shock wave runs up the pipeline, as
the fact of the closure of the valve is communicated to the pipeline liquid, in order to stop it. The
effect can repeat cyclically at certain values of valve opening in long pipelines, a highly
undesirable situation. Water hammer is most easily prevented by ensuring the valve opening
time is long compared to the characteristic time constant of the pipeline, given by
where L is the pipe length and v is the shock wave velocity, which is dependent on both pipe and
liquefied gas properties but is quite typically around
GAS TECHNOLOGY 203
Figure 2.49 Gas pulsation filter and its electrical analogue.
The connection of pumps in parallel has other problems, as when one of
the parallel units suffers a breakdown and may be driven in reverse by the
others. Best practice here is to install a check-valve in each of the pump
outlets. Pulsations caused by pumps are common and may be dealt with
using fluid circuits analogous to the d.c. power supply filtering circuits used
in electrical engineering. The classic d.c. power filter supply circuit has one
or more series resistors with capacitors in parallel. An analogous circuit in
gas plumbing comprises restrictions such as orifices or small diameter pipe
sections interspersed with larger sections of pipe or small pressure vessels in
series (Figure 2.49).
2.4.2 Valves and other pipeline components for gases
Transistors are often compared to valves controlling gas flow in their ability
to control the flow of electric current. However, gas valves are intrinsically
much more imperfect in their abilities with respect to gases than transistors
are with respect to electricity. Valves often suffer from some or all of the
following problems:
flow restriction
leakage across seat
leakage to outside
high actuation force
nonlinear flow versus actuator position
wear and short life
trapping of particles leading to leaks, wear and failure
slow actuation time.

The results of emphasis on different aspects of these problems is that many
different forms of valve are available, which vary enormously in cost and
complexity (Figure 2.50).
Figure 2.50 Some common types of valves: (a) poppet; (b) ball; (c) fluidic diode; (d) standard
spring check/relief.
Poppet valves should be considered for many plant designs where
reliability and frequent valve actuation is required. They can be made very
reliable and offer a high resistance to leaks under high differential pressure.
Only a very small movement is required for actuation, bellows or
diaphragms can be used to prevent leakage to the actuating stem, and sliding
contact can be designed out almost completely, giving low wear rates.
Poppet valves can also be designed for very rapid actuation in safety
applications.
27
However, they are normally placed so that increasing
differential pressure leads to increasing actuation force. They also offer a
relatively high resistance to flow and do not normally make particularly good
metering valves.
Valves such as needle valves or gate valves can be designed so that their
27
The actuation speed of a valve is dependent on the power driving it and its size in general.
One project the author was involved with required ultrarapid pulsation of an oxygen supply.
The most effective controllable high-speed valve was an electronically driven solenoid poppet
valve. This could switch from open to closed in a few milliseconds by driving it with more
electrical power than normal and by driving it down as well as up, rather than relying on the
return spring. However, for a system requiring a continuous stream of pulses, a rotary valve
driven by a high-speed electric motor proved to be a better system. This system was capable of
over 15 000 pulses per minute. A special design of valve with many ports (rather like the
mechanism of a siren sounder) could have raised this high frequency even further. However,
with regular pulses of gas there are severe problems with resonating effects and also a loss of
pulse amplitude because of smoothing effects in the pipework.
GAS TECHNOLOGY 205
Figure 2.51 Examples of valve opening characteristics.
effective orifice area is a linear function of actuation. However, the rate of
flow on opening a valve is rarely in practice a linear function of the opening
of the valve because of the effects of the pipeline in which they are placed.
Other valves, such as ball-valves, have a nonlinear slow-opening character-
istic, while still others, such as butterfly valves, have a nonlinear fast-
opening characteristic (Figure 2.51).
An approximate idea of the nonlinearity expected in the flow out of a
valve with upstream and downstream pressures and P can be obtained by
noting that for small opening values the valve will typically behave as a
critical orifice, like a small orifice plate in a large pipe, at least for supply
pressures much greater than downstream pressure. In this case, the flow rate
expected will be proportional to the valve area of opening (A) and the
absolute supply pressure
As the valve approaches the bore of the pipe in opening area, a low
pressure-drop orifice formula is appropriate
where P is the downstream pressure. However, for most valves, the flow at
larger opening angles is distorted by drop in the supply pressure and by flow
resistance in other components.
The effect of the surrounding pipework can be understood if a combi-
nation of orifices and linear flow elements in series with the valve is imagined
(Figure 2.52). The effect is that most valves tend to have a fast-opening
characteristic in practice when fitted in a pipeline. As a result valves having a
slow-opening characteristic are favoured if it is required to provide better
control of low flow rates.
Figure 2.52 Flow versus pressure for a valve in a pipeline.
Non-return/check valves. Non-return (NR) or check valves are an essential
feature of many pipeline systems. They are the fluid equivalent of the
electrical diode, conducting gas in one flow direction but not the other. They
serve to prevent or restrict reverse flow in many operational situations; for
example pump outlets and PSA outlets. NR valves are frequently required
for safety reasons, too, as in gas bubbling systems, where the NR valve
prevents the process liquid entering the gas line.
It is possible to make devices with no moving parts, sometimes dubbed
fluidic diodes, which have some of the function of an NR valve. These rely
on the inertia of the fluid (most often liquid but gas diodes are possible) in a
swirl chamber. In the forward direction, the fluid flows smoothly into the
centre of the chamber and out of the periphery. In the reverse direction, the
fluid flows into the periphery and sets up centrifugal forces which greatly
increase the reverse-flow resistance.
Seals. Many gas components cannot be entirely constructed out of the
most ideal material solid metal. Although all-metal valves and pumps are
possible, and some valves used for high purity semiconductor gases are
all-metal, most practical devices require rubber or plastic compliant and
sometimes sliding surfaces to provide a gas-tight seal.
Distinction should be made between seals which are static and subject
only to the small, slow movements associated with differential thermal
expansion, and those which are required to undergo more rapid com-
pression or stretching or are, worse, subject to sliding. Static seals may be
made of materials which would be useless for dynamic seals, materials which
would show little protection against abrasion, for example, or which swell.
Distinction should also be drawn between seals which are at ordinary
pressures and those at such a high pressure that they can be destroyed by
GAS TECHNOLOGY 207
being extruded out of the volume in which they are confined. At pressures
above 100 bar it is often still possible for ordinary O-ring seals of common
composition to be used. However, at such high pressures it is necessary to
use them with techniques such as backing rings, which are metal rings
placed so as to prevent O-ring extrusion. The backing rings are on
cylindrical seals and are often wedge-shaped in cross-section (mitre rings).
They function by being rammed by the operational pressure against another
ring or a shoulder on the cylinder to form a very close fit with the bore of the
cylinder, allowing no gap for extrusion of the polymer seal ring. Alterna-
tively, specially shaped soft metal seals, such as the wave ring or delta
ring, can be used, or a direct metal-metal sealing assembly used, such as a
bull-nose fitting. With careful attention to detail, good sealing up to 10 kbar
or more is possible (Anon, 1975).
Although not difficult compared to high-pressure devices, even modest
pressure seals require some design work. For packed or stuffed seals, the
basic principle is that the pressure exerted by the seal on the surfaces must
exceed the fluid differential pressure applied. If this is not the case, then the
packing will be pushed back by the fluid and leaking will occur. The
necessary pressure is often provided for simply by allowing an adjustable
pressure on the stuffing via a gland nut, which is pretightened to a known
torque, or simply tightened until leaking stops to an adequate degree. Often
a lubricant is included in the composition of stuffing seals used in gas
apparatus, as lubrication from a minute leak, used when containing
pressurised oil, is not possible.
There is both general published and manufacturers data on the required
size and form of grooves for O-rings. These published guides achieve a
precompression of the seal which is sufficient to guarantee no leaks under
usual operating conditions. Seals for pistons or rotating shafts, such as
sprung lip seals, which must both seal and not impose a significant torque or
heat up are more difficult to design but specialist suppliers can advise on
suitability and fitting (section 2.1.3).
Elastomers (rubbers) are high molecular weight polymers (Table 2.10).
Their molecules are held together by a relatively small number of primary
bonds (cross-links) between polymer chains, and a large number of
secondary bonds, which are much weaker links. It is the latter links which
are stretched or broken during stretching of rubbers and reformed when the
applied force is relaxed. The primary bonds ensure that the rubber article
returns to its former shape when force is removed and they are what
distinguish an elastomer from an ordinary plastic.
Swelling is the natural reaction of any elastomer when its secondary
bonding breaks down under the influence of a solvent. The solvent, which
would in the absence of the elastomers cross-links dissolve it completely, is
absorbed into the material and destroys the secondary bonds to an extent
depending on potency. The result is a swollen elastomer. Swelling may be
slight, reversible and associated with only slightly degraded elastomeric
properties. Frequently, however, swelling is gross, 100% or more increase
in dimensions is possible, and associated with undesirable degradation such
as stickiness or crumbling.
Seals are a compromise between the following issues:
tight gas seal versus high frictional forces;
elasticity versus slipperiness;
elasticity versus solvent resistance;
cost versus lifetime

There is no perfect seal material. Swelling on exposure to the wrong
solvent, as noted above, leads to problems. Swollen seals often suffer
further knock-on problems caused by swelling such as frictional heating or
abrasion leading to failure. Many elastomers will lose elasticity and fail as a
result of oxidation in air or oxidants. Many plastic materials suffer creep,
i.e. permanent deformation, which frequently leads to loss of gas-sealing
capability if not outright failure. Cryogenic service has even more problems,
with most polymers becoming brittle long before liquid nitrogen tempera-
tures. A graphic demonstration of this is easily achieved by dipping a rubber
item in liquid nitrogen and hitting it cautiously with a hammer; even the
toughest rubbers generally shatter with little provocation.
Grades of elastomers vary enormously in their actual performance and it
GAS TECHNOLOGY 209
is nearly always worth contacting the manufacturer for specific information.
Also, many practical elastomers are blends of several different compounds
and may vary in their degree of cross-linking, varying their performance
over wide limits. Highly cross-linked polyurethanes, for example, are more
resistant to chemical attack than low cross-linked polymers with a similar
chemical make-up. Fillers vary the property of the basic polymer in practice,
too. Properties which may be important in practice, such as wear resistance
and coefficient of friction, are affected by the addition of fillers, such as
carbon black, and additives, such as rosin, respectively. Again the polymer
manufacturer (or the firm which carried out the compounding and
processing) should be able to help.
One method of sealing joints in gas systems is also a means of lubrication.
Threaded joints in gas systems, which rely on a tapered thread to seal them,
although not in general the first choice for a new system, are completely
satisfactory if correctly assembled. The threads, particularly of stainless
steel components, will not in general tighten satisfactorily without a
lubricant grease; they have a tendency to galling, in which small areas of
the steel faces cold-weld to each other and are then torn apart leaving a
damaged finish.
The choice of lubricant is very restricted (Figure 2.52) and, in fact, the
leak tightness of the joint can be improved by the use of a thread sealant.
These both provide lubrication and fi l l in any minute gaps left in the
assembled taper joint. For many purposes PTFE tape is the best choice for
thread sealant. It is compatible with a large range of gases and is satisfactory
at surprisingly high pressures.
Gas compatibility. Gas companies all issue gas compatibility tables, giving
metals and seal/hose materials suitability for each gas (Table 2.11). Note
that fluorine gas can only be used in systems that have been passivated for
fluorine service (section 3.5.3). As described in section 2.5.9, hydrocarbon
grease and oil can form an explosive when oxygen is dissolved into them at
high pressure. Even apparently non-inflammable materials, including, for
example, chloro and fluoro plastics and rubbers, are, in high pressure pure
oxygen, highly inflammable or worse.
Oxygen compatibility is tested by placing a sample in a bomb (a small
high pressure vessel), applying high pressure oxygen and then attempting
ignition electrically. The temperature can be steadily increased until a point
is reached at which the test compound sustains combustion and ignites. In
pure oxygen, many ordinarily stable materials are unstable and burn easily
with very low autoignition temperatures. Some examples are given in Table
2.12.
Lubrication. Whilst for many gases ordinary mineral oils and greases give
satisfactory service in gas equipment, for some gases, and especially at high
GAS TECHNOLOGY 211
pressures, special compounds are necessary. Oxygen service, and service
with oxidant gases, are generally the most stringent. Lubricants should be
certified or tested for ignition temperature before use.
Corrosive gases, such as hydrogen chloride, are not in general a problem
with lubricants and seals, unless, like, for example, chlorine or fluorine, they
are also oxidising agents. Fluorine is exceptionally difficult to work with,
requiring seals and lubricants with a high state of fluorination such as the
perfluoropolyethers. For many purposes, the perfluoropolyethers, such as
Fomblin, can be used. These are made in a range of molecular weights
from light lubricating oils to greases. They are fully fluorinated and
exceptionally stable.
A typical polyether might have the formula
These compounds have a lower viscosity than their molecular weight would
suggest. A compound of molecular weight of 2600 a.m.u. has a pour point of
42C, for example. (The pour point is 5C above that at which the oil begins
to set into a jelly or wax.) They are water repellent and have the curious
property of being able to dissolve surprisingly large amounts of oxygen; 50
or 100 ml liquid is not untypical, a fact which may have applications in
medicine (section 3.9.1).
The perfluoropolyethers are, however, rather poor in their lubricating
properties. They have a low surface tension and a very high viscosity index.
This means that their viscosity decreases more steeply with increase in
temperature than ordinary oils. Machinery using perfluoropolyethers is thus
more susceptible to seizing because an increase in temperature can lead to a
decrease in lubrication, leading to further increase in temperature and so on
until failure. They are thus suitable only where conditions (gas pressure,
bearing pressure, relative speed of bearing surfaces) are not too arduous.
Unlubricated bearings can be used in many gas applications at low speeds.
Bearings such as Glacier DQ, which combine PTFE self-lubricating plastic
with graphite and soft metals, are particularly suitable as they are
substantially non-reactive and can also be used at cryogenic temperatures
satisfactorily, when normal lubricants such as oils are frozen and ineffective.
Figure 2.53 (a) Pressure regulator, (b) Output from a single-stage pressure regulator versus
flow and inlet pressure.
Regulators. There are many different types of gas pressure and flow
regulators. However, the term regulator is normally reserved for poppet
valves connected to a control diaphragm, as in Figure 2.53a (a practical
embodiment is shown in Figure 3.22). The spring can be adjusted in
compression with the screw to adjust the outlet pressure or changed for a
different one for different pressure ranges. This simple mechanical combi-
nation has been the basis of gas pressure regulation almost since the
conception of high pressure gas. A standard regulator is simple, inexpensive
and reliable. However, it has all the problems of a poppet valve, together
with the problem of the diaphragm, which can leak. In addition, the pressure
regulator performs its function only imperfectly, as is shown in Figure 2.53b.
Gas regulators, finally, can also lead to pressure instabilities in systems,
i.e. oscillations and pulsations, which can be very detrimental to a system's
function and lifetime.
Gas regulator theory. The ideal gas regulator would employ a very large
diaphragm to operate a very small poppet valve and would be opened by a
very long spring. It would open very wide until the downstream pressure
GAS TECHNOLOGY 213
approached the set point given by the spring pressure, when it would rapidly
close down to the correct orifice for the flow rate needed at that pressure.
The dynamic theory of gas regulators is complex, with the inertia of the
moving parts interacting with acoustic resonation of the gas in the pipeline.
However, the static theory of pressure regulators is simple and still useful. A
standard regulator, with diaphragm area A and poppet valve area B tends to
have an outlet pressure which rises with fall in inlet pressure and falls with flow
rate as follows
where the outlet, inlet and atmospheric pressures are P
o
, P
i
and P
a
, the spring
exerts force where x is its displacement, and the constant in the
orifice flow versus pressure and displacement curve is k, and the flow rate is Q.
A two-stage regulator is much better than a single-stage because the inlet
pressure is stabilised, reducing to immeasurably small (0.1 psi or so in typical
designs) the variation in due to changes.
Other devices which can be useful in pipeline design are:
The FLOSTAT: a device which tries to maintain constant outlet flow,

rather than outlet pressure;
the excess flow shut-off: used for safety purposes, this shuts off when flow
exceeds a preset l i mi t , and can only be reset by manual intervention;
flashback arrestors, which prevent flames travelling back along pipelines

carrying inflammable or other reactive gases. These are valves which shut
off if an excessive pressure rise or backflow occurs. They also contain a
metal grid snuffer assembly, which cools a propagating flame and
extinguishes it. Some are designed to autoreset, others require a manual
reset. They are mandatory with some inflammable gases such as acetylene.
2.4.3 Removing particulates: filters and cyclones
In most gases, particles above a few microns are eliminated during production
or purification. For many processes, this is sufficient; a few micron-sized
particles, for example, are unlikely to affect the functioning of a steelmaking
process. However, there are many more processes where particulates even at
the submicron level are unacceptable.
The vital parameters of a filter from a users point of view are:
its efficiency in removing particles versus particle size;

its pressure drop;
its compatibility with gas;

its re-entrainment;

that any possible contamination, gaseous or particulate, caused by the filter
itself must be acceptable;

susceptibility to clogging or other lifetime limitations;

its cost.
Many of these parameters interact. Filter drop and efficiency are typically
complementary, i.e. very efficient filters will have a greater pressure drop
than less efficient ones. Filters amount to a finely divided form of material
which can be expected to deteriorate much more rapidly than normal bulk
material if there is any degree of degradation reaction between filter and gas.
Compatibility of filter material is thus very important. For example, any
filters used with oxygen gas must be certified as suitable for that service;
many organic filter materials form a serious fire hazard when filled with
high-pressure oxygen gas.
A filter is superficially a very simple component. It is not, however, simply
a sieve on a microscopic scale. It might be thought that if a particle is smaller
than the sieve size, it will penetrate, while if it is bigger, it will not. However,
this is not the case. As the size of a sieve is shrunk to a few microns or less, its
functioning changes. It is possible to produce a submicron filter which
functions as a sieve but this is most unusual practice .
28
Most submicron filters
work as sieves for the very largest particles but rely on other mechanisms for
medium-sized and very small particles.
Macroscopic diffusion is one process, whereby small particles, so small
that they are subject to Brownian motion, wander into small interstices of
the filter and are then permanently captured by short-range intermolecular
forces such as van der Waals. The other important mechanism is that of
inertial impaction. Larger particles, which are still smaller than the larger
holes in the filter and might therefore penetrate it, are retained by impacting
into the filter medium as they pass through the filter. Once impacted, the
particles remain trapped by electrostatic and van der Waals forces.
Submicron filter action: most penetrating size. A useful concept in
assessing inertial impaction on filters is the Bremstrecke or stopping
distance. This is the distance which a moving particle takes to slow down to
essentially zero velocity in a stationary medium. For a particle radius r,
density travelling initially at velocity in a fluid of viscosity
which stems directly from the integration of Stokes law.
Next consider that a particle must be swept around a filter fibre of
diameter R at about the stream velocity If the stopping distance of the
particle is large compared with R, then it will be impacted and the particle
will carry on in a straight line rather than following the stream
28
Nuclepore
TM
filters are sieving filters at the submicron level. They actually consist of a plastic
membrane with accurate diameter non-overlapping holes randomly etched into them. They
have a very high pressure drop relative to filters of the same rating, are expensive, easily
overloaded and relatively fragile. They are used mostly to collect samples of particulate matter
for examination under an optical or electron microscope. How they are produced is quite
interesting. Blank plastic membranes are heavily irradiated in a nuclear reactor or other
facility. The ionising particles leave the polymer of the membrane with some chemical bond
damage but only in the region where the particles passed through. These damaged regions are
attacked much more swiftly by a strong chemical etch bath, which t hus drills a pattern of holes
through the membrane, all the holes being the same and of a size determined by the etching
time.
GAS TECHNOLOGY 215
Figure 2.54 Filter efficiency versus particle size, (a) Filtration mechanisms; (b) penetration in
practical filter.
i.e.
Thus will be obtained. This is the critical size below which the particles
will not impact but will be carried along streamlines through the filter, in the
absence of other effects. Typical filters have critical sizes in the hundreds of
microns to 0.5 or so for impaction.
At very small sizes, the diffusion capture process operates. Here the
Brownian motion scale must exceed the filter interstices typical size, in order
that in the time allowed during passage through the filter, the particle can
diffuse to a solid surface and stick. The distance x gone by a particle under
Brownian motion is proportional to the square root of the time allowed, as in
other diffusion problems. For particles bigger than 60 nm or so (the mean
free path for nitrogen at NTP), the Einstein formula gives (Einstein, 1924)
Setting and x as R gives
Typical filters give in the order of a fraction of a micron for below which
particles are captured.
Between the almost perfect retention of very small particles by diffusion
and good action in the sieving and impaction size ranges, there is for most
filters a size for which the filter is least effective (Figure 2.54). The most
penetrating size will vary, as will the efficiency or penetration coefficient at
that size. However, for example, a certain electronic gas filter has a most
penetrating size of about 0.3 at which penetration is 0.0001%.
Filter materials. The majority of filters are made of some kind of fibre.
Filters can be made of thick beds of material or can be membrane filters,
where the bed can be as little as a few tens or hundreds of microns thick.
Fibrous filter cloths are woven or felted from materials as diverse as glass
fibre and Teflon, whilst other filters rely on porous ceramic or partially
sintered metal powder compacts. The latter types, typically made from
alumina, stainless steel or nickel are discussed in section 3.5.5. They are
expensive and used almost exclusively in wafer fabrication plants. A filter
cloth or membrane is often given a face velocity rating. This is the maximum
gas velocity through the medium that will not result in entrainment on the
outlet or, worse still, break-up of the filter material.
In filters where some clogging might be anticipated, two or more different
filters, of varying degrees of fineness, are laminated together or a continuous
variation in fineness throughout the thickness of the filter is produced. The
coarse filter layers are on the inlet side and the fine filter layers are on the
outlet side so that clogging and re-entrainment are minimised.
Gases are generally very clean when delivered from production plant or
from bulk or cylinder sources; many orders of magnitude cleaner than the
several mg loading commonly seen in air. Gas line filters are therefore
by no means always necessary although a coarse-sintered filter can stop
occasional large particles, for example swarf from machined components,
from getting to the point of use. Gas filters can in any case be designed for
minimal loading, while often a pressure drop of a fraction of a bar is not
significant in a system. As a result gas-line filters are often very small units.
By contrast air filtration is almost inevitably required on the air inlet of gas
production and in gas applications needing air, and furthermore both
loading and pressure drop are considerations. Fibreglass filters are often
used, with fibres of around 2 or 4 supporting much finer fibres in a
paper-like matrix held together with a small proportion of adhesive. A very
GAS TECHNOLOGY 217
Figure 2.55 Cyclone dust separators.
large area of this filter mat is then folded up into a compact module which
offers a very low pressure drop (often measured in inches in a water gauge)
and long intervals between replacement. Similar HEPA (high efficiency
particulate air) filters are mounted in the roof and used for cleanroom air in
wafer fabrication plants.
Occasionally, in dusty parts of the world, it may be necessary to employ
coated or viscous air filters. In these the fibres of the filter are coated with a
heavy oil to which particulates adhere strongly and will not re-entrain. In
these types a relatively coarse filter can still be effective and the dust loading
they can withstand is higher than that of uncoated dry filters.
Cyclones. Cyclones are devices that can clean gases via a centrifugal
separation action (Figure 2.55). They are rarely used in industrial gas
applications because these mainly use clean sources and cyclones are mainly
applicable to gases with a high burden of relatively course particulate
contaminant. However, cyclones do find application in exhaust treatment in
many gas applications, as well as in some gas production processes. They are
capable of separating out dust as fine as or so. The most common type
is the conventional reverse-flow type, whilst straight-through cyclones with
fixed or moving impellers are rather unusual.
The balance between Stokes law drag on a particle and the centrifugal
force to which it is subjected in the cyclone is the essential nub of cyclone
theory. A conventional cyclone analysis leads to an inverse square-root
dependence of critical particle size on gas velocity and cyclone size, i.e.
Cyclones normally have a much lower pressure drop than any workable
filter design. The pressure drop is approximately proportional to gas flow
squared in the operating range of many designs. However, cyclones have little
capability for very fine dusts; the minimum particle collected is typically in the
diameter region. Another problem is that the particulate receiver
must be sealed, otherwise large flows around the receiver may cause
re-entrainment and loss of efficiency. It is interesting to note that cyclones
have a fluidic diode effect (section 2.4.2), i.e. they have a much higher
pressure drop in the forward flow direction than in the reverse direction.
A typical cyclone is used for processing waste gases; perhaps ensuring
compliance with dust emission regulations or recovering by-products in the
collected dust. The dimensions of cyclones are critical design parameters and
must be accurately calculated in practice. Typical proportions might be 1.4 m
height, 300 mm throat diameter, 200 mm 70 mm inlet, 70 mm diameter gas
outlet. In such a typical cyclone, with an inlet gas velocity of the
cut-off size expected would be about diameter at a dust density of around
It is now possible to model cyclone action with simple computer models,
which are now used to predict performance and enhance design. These
models have made cyclone design more scientific and, incidentally, have
shown clearly how small, apparently innocent, design changes in the cyclone
shape can have significant effects. Even as small a change as a dished end on
the top, useful for strength in a pressurised system, can change cyclone
performance markedly.
Electrostatic precipitation. Electrostatic precipitation involves electrically
charging particulates and then using electric forces to attract them to a plate
where they are discharged and collected. Electrostatic precipitators are more
useful in cleaning up dirty gases (section 2.5.14). They can have very much
lower pressure drops than filters or cyclones, whilst still removing large
amounts of particulate (Brauer and Varma, 1981).
The simplest precipitators apply a negative voltage of 10 kV or more to a
wire electrode, which may simply be a wire of small diameter or a more
substantial tape or rod with sharp points at intervals along its length. The
corona discharge produces electrons and ionises the gas, and some of the
mainly negative gas ions produced transfer electrons to particulates. The
negatively charged particulates are then attracted to the positive collection
electrode. Vibration, either continuous or regular impact rapping, must be
applied to ensure that the precipitator does not get too thickly coated in dust
and to collect the dust.
2.4.4 Gas flow control systems
Simple gas panels. The standard paradigm for control systems must today
be electrically actuated valves controlled by microprocessor-based digital
GAS TECHNOLOGY 219
electronic logic circuits known as programmable logic controllers (PLCs).
However, for a modest gas control system (perhaps only two gases need to be
supplied, one for purging, one for running, at constant pressures) a simple set
of manually operated valves and regulators, together with some ancillary
components, will often suffice. Where the component sizes are conveniently
small (say below 30mm pipe diameter), the components are often mounted
on a metal board as a separate subsystem built in a factory rather than
assembled on-site. Such subsystems are often called gas panels. Cylinder
gases will typically use pipe of 6 mm external diameter so gas panels are
popular for compressed gas service.
Gas panels are often required to govern the ratio of gas flows so that the
mixture of gases delivered is of constant composition, despite changes in the
total gas flow. This can be achieved by regulators with critical orifices or
needle valves, or, more accurately, by mechanical flow ratio regulators using
balanced diaphragms. Alternatively, and increasingly today, two or more
thermal mass flow controllers are used. These can be used to govern flow rates
separately or, with an electronic link between them, to maintain constant
ratio. Filters are often mounted on gas panels, together with non-return
valves to prevent backflow and possible contamination of the bulk gas store.
In the case of cylinder gases, the input to the panel is via a pigtail or the more
expensive Teflon-lined hose, with the cylinder connection compatible with
the chosen gas type on the end.
Gas cabinets. It is preferred on safety grounds to place toxic and corrosive
gases outside in a protected area but one in which ventilation to the outside air
is free and plentiful. However, this is not always practicable. Cleanliness
requirements work against this simple solution, as do production site layouts
where equipment using gases is often placed far away from outside walls. It is
for these reasons that toxic or corrosive gases in cylinders are most often
handled by a gas panel inside a gas cabinet. This is a metal cupboard with
space for two cylinders of gas (one active, one purge gas) and the gas control
panel.
Gas cabinets are strong, of coated 1 or 2 mm steel construction, fitted with
an extractor fan which sucks air (and any process gas which may have leaked)
out and off to a vent stack or, where quantities may be significantly large, to a
scrubber system which will remove the dangerous gas. Purging of the pigtail is
particularly important for these gases, which otherwise give out a puff of a few
hundred of gas each time the cylinder is changed, which is unpleasant and
potentially unhealthy for the operator. Good purging is achieved by
evacuating the pigtail using a blast of purge gas through a venturi (this can give
100 mbar or so vacuum, enough to improve the purge cycle radically over
simple blow-down) then repressurising with inert purge gas and repeating the
process several times. The purge gas can also be used to purge out the
downstream pipeline ready for maintenance operations, for example.
There are generally two doors on a gas cabinet, one to allow limited access
for adjusting one of the gas cabinet pressure regulators or valves, and the
other, larger, door for the addition or removal of cylinders. Figure 3.23
shows a typical compressed gas handling cabinet, showing two process gas
cylinders and one purge gas cylinder for high purity gas service in the
semiconductor industry.
Electronic flow controls. For larger systems electronic controls are justi-
fied. Some will have a complete PC or similar small computer, with display
and keyboard, as their master controller. A plethora of software and
hardware is aimed at satisfying this market and prices have tumbled in recent
years. In many markets inertia forces are great but electronic controls will
eventually take over from pneumatics almost completely.
For variable flow rates the valves are controlled by sensor feedback loops
to proportional/integral/differential (PID) controllers whose set points are
in turn set by programmable logic controllers (PLCs). For on/off valve
control electrical signals direct from PLCs are used. Sensors are typically
flow sensors or pressure sensors with less often (cryogenic) level detectors,
thermocouples or more exotic probes being used. Mass flow controllers
(MFCs) are the mainstay of applications such as semiconductor manufac-
ture where many fairly small flow rates must be accurately controlled. The
MFCs in semiconductor practice are usually directly connected to the
computer that controls the rest of the equipment, rather than having their
own PLC.
Despite the predominance of electronic control, however, there are
occasions where an alternative approach may be justified even on new
equipment.
Pneumatic controls. As stated, conventional control systems today often
use electronic logic to decide which valve should be actuated and by how
much. However, it is often inefficient to try to use electronic valves, such as
solenoid valves, directly on process gas flows; large solenoids drawing
significant amounts of power would often be required. Pilot-operated
pneumatic valves are therefore used for higher flow applications. Small
electric solenoid valves are the interface between the electronic world and
the gaseous world. These pilot valves switch a compressed air supply to the
actual process valves, which are powered by pneumatic piston or bellows
working off a gas or compressed air supply rather than electricity. There may
even be two stages of pneumatic actuation involved in large process plant,
the second stage being the solenoid switching pneumatic valves which in turn
switch compressed air onto the process valve actuator.
A pure pneumatic control system is simpler in concept than these more
typical systems because it uses compressed air not just for the actuation of
the process valves but also for the logic and amplification normally done
electronically. Given the low cost and versatility of electronic controls, it
GAS TECHNOLOGY 221
may be doubted that pneumatic controls have any application whatever and
it is certainly true that they are now rather unusual. Pneumatic controls,
however, have a number of advantages over conventional general purpose
electronic controls, at least for small systems. For example, they:
are intrinsically safe, no sparks being generated in normal pneumatic

systems;
can sometimes function without an external power source (if the process
gas itself can be used instead of air in the control system);

are not subject to electromagnetic interference;
do not radiate electromagnetic signals which can interfere with other

controls;
are capable of functioning in dirty, hot and environments with suitable

filters;
are often less expensive than similarly reliable electrically actuated valves;
allow relatively high actuation forces to be easily obtained without

compromising on actuator speed.
Against these advantages are considerable disadvantages where complex
functions are concerned. It is not easy to generate a complex sequence of
actions with pneumatic controls; these sorts of sequences are much more
easily carried out by the electronic programmable logic controller units
described above. One might add a further disadvantage: that the younger
engineering designer today may well be rather unfamiliar with pneumatics.
Fluidic controls. In principle the terms fluidic and pneumatic should be
synonymous, since fluidic components invariably employ air or gases as the
working fluid, just as pneumatic components do. The term fluidic is,
however, now reserved for systems which employ a set of non-moving part
pneumatic components; the fluidic diode mentioned in section 2.4.2 is an
example. A simple example of a fluidic active component is the commonly
used wall-attachment (or Coanda) effect bistable (Figure 2.56). This device
has a compressed air input, two signal inputs and two outputs. When the
input tube at the active output is puffed momentarily on, the input stream is
switched to the opposite output. These devices can easily be made an inch or
so in size and a fraction of an inch thick. Their size is usually small compared
to the interconnections and pipework in a typical system. The fact that
fluidic devices have no moving parts is ingenious and makes for high
reliability. Fluidic devices are efficient devices for switching flows, switching
at high speed (ms) and with high amplification factor (power controlled/
power controlling).
Analogues of most electronic components can be produced in fluidic
form. Amplifiers, OR and AND gates, timers and oscillators are all
possible. Essentially, fluidic controls are applicable where pneumatic
controls are appropriate, especially in situations which value the advantages
Figure 2.56 Wall-attachment fl ui di c amplifier.
which stem from the absence of moving parts, i.e. resistance to vibration,
long lifetime, and an absence of a requirement for lubrication. The major
disadvantages of fluidics today, however, are their poor availability and lack
of support from suppliers.
Although fluidic systems are unusual, there are certain useful fluidic
subsystems which are more commonly seen; for example the fluidic
oscillator molecular weight sensor. These are made using fluidic switching
technology and achieve good precision. In effect they are the same in
principle as an organ pipe runni ng off a carefully regulated gas supply whose
pitch varies with the square root of gas molecular weight. The frequency of
oscillation can be easily converted into a readout of average molecular
weight.
Back diffusion. Back diffusion refers to the following situation. Suppose
that the exhaust end of a gas system is open to the air and air is an
undesirable impurity in the system. If the exhaust flow is low, or the exhaust
pipe too short, then it is possible for small amounts of air to diffuse back up
the exhaust, against the general gas flow, and spoil the process.
Why this arises is as follows. Each molecule of air and gas is moving at
around 400 m per second but, of course, it soon hits another molecule
and this changes its direction. In this way, it progresses according to the
diffusion equation described in section 1.3.10, rather than linearly forward
at 400 m In fact, for the average molecule to go any reasonable distance
(section 1.3.10), the time allowed must be very long. However, a tiny
fraction of the fastest molecules can travel quite a long way.
If back-diffusion occurs in a pipe with plug flow, then the concentration of
back-diffusant rapidly falls exponentially (exp(x)) to zero. However, if
back-diffusion occurs in a pipe with viscous flow, then wall creep occurs
GAS TECHNOLOGY 223
and the back-diffusant falls only hyperbolically (1/x). (It is well worth
remembering, too, as noted in section 1.3.10, that diffusion as a phenomena
is pressure-sensitive; back-diffusion will be more serious in low-pressure
systems.) Back-diffusion effects seen in practice are between the exponen-
tial and hyperbolic in effect.
Leaks in pipework. Finding leaks in completed pipework is an essential
activity both during original installation and after any modifications. If the
pipeline can be pressurised with an inert gas such as nitrogen, then simple
tests, variants of the time-honoured soapy water, can be employed. The use
of synthetic surfactants or non-water-based bubbling solutions can occasion-
ally be useful. Bubbles as small as a few millimetres across forming in a
minute can easily be seen, corresponding to leak rates of the order of
mbar , which is a fairly small leak rate; a standard (200 barg,
501 water capacity) cylinder would be exhausted only after hundreds of
years.
Helium, being a smaller and faster-moving molecule than nitrogen, will
find smaller leaks with soapy water. Furthermore, helium can be used with a
helium leak detector.
Helium leak detectors are basically simple mass spectrometers. They are
tuned to helium, which is a rare gas (< 1 ppm) in the atmosphere, giving a
small background signal. They can be used in several ways but the most
effective way is as follows.
1. The pipework is evacuated and connected to the He leak detector.
2. A small cylinder of He is used to play a stream of He on the outside of the
pipework via a hypodermic needle.
3. When He is played onto the pipe at a leak, He atoms are almost
instantaneously sucked down the pipe towards the leak detector vacuum
pump and detected.
4. The needle source of He is used to pinpoint the leak precisely.
In this way, leaks of the order mbar can be detected, and often
located to within millimetres.
Many pipework systems cannot be evacuated; they may have seals which
only seat correctly if under positive pressure or they may include vessels.
Many large vessels, even those designed for a several bar overpressure on
the inside, may fail by buckling under atmospheric pressure on the outside if
subjected to vacuum. If the pipework cannot be evacuated, then a different
type of leak detection strategy is called for. Pressurising the system with an
easily detectable gas is the way normally employed; for larger leaks pressure
drop can be observed.
Helium sniffing of the air around the vessel, using a mass spectrometric
detector as used in the vacuum leak detector but adapted to take an
atmospheric pressure sample, is possible. However, a cheaper instrument,
relying on the detection of a halogenous gas, typically a CFC or or one of
these diluted with nitrogen, using a sensitive electron capture detector is an
alternative which will suffice for many applications. Even less sensitive than
this are simple thermal conductivity-based leak detectors, which, however,
have the advantage that they will detect any escape of gas with a thermal
conductivity differing from that of air; no sniffing of gas is necessary
(although sniffing gas which is very high or low in thermal conductivity will
improve the sensitivity of the method).
Probably of similar sensitivity is the technique of listening for leaks. In this
a microphone tuned to receive ultrasound (typically around 40kHz, at
which frequency inexpensive transducers are available) is used as the sniffer.
When the microphone is placed near a leak from a gas source at high
pressure (over 1 barg or so), it receives the small ultrasonic noise signal
caused by the turbulence of the escaping gas. This is amplified and operates
an audible signal and visible analog meter for the operator. Low-pressure
systems can be checked without applying any differential pressure by placing
a 40 kHz transmitter inside the vessel or pipework and probing the outside
with the microphone.
Repairing gas leaks. Although the pressures used in cylinder gas systems
may seem high, up to 300 bar, and the size of bulk gas equipment large, the
principal sorts of leaks seen, and the ways to repair them, are not dissimilar
to that of any pipeline system. Before attempting any repair, it is always best
to shut off the gas supply and, if a dangerous gas is in the line, purge with
nitrogen or other inert gas.
High-pressure fittings of the Swagelok type can often be sealed by
simply fastening them up tighter, if possible after checking for dirt on the
seal faces. However, overtightening the fittings will bell out the face
opposite the pipe olive and make it difficult to dismantle the equipment.
High-pressure fittings with face seals of the VCR type sometimes require
a new seal washer to be fitted. This type of fitting should always be fitted with
a new seal washer after dismantling; they may leak if a new washer is not
fitted.
Valve stems, after time or extensive use, will tend to leak because their
seal gland packing is no longer sufficiently tight. A modest tightening of the
gland nut will normally effect a seal. The ease with which the valve may be
operated should be checked after this, as an overtightened gland can make
the stem difficult to turn. Cylinder valve outlets are often subject to leaks. It
is best to turn off at the cylinder valve and look for a cause rather than
tightening the j oi nt past its intended torque. Typically, the cause will be dirt
on the mating faces or a worn sealing gasket, if one is fitted. If a cylinder
valve will not seal shut, it may be possible to fit a gaslight metal plug or cap in
the outlet. (The plastic dust plugs and caps provided with some cylinders to
keep dirt out of the cylinder outlet are, needless to say, not sufficient.)
GAS TECHNOLOGY 225
2.4.5 Liquid gas storage and handling
Liquefiers. In former times, the demand for oxygen in gaseous form was so
large that the small amounts of liquid nitrogen needed could easily be drawn
off a large air-distillation plant without affecting its performance. However,
the relative demands for oxygen and nitrogen have reversed and much of the
nitrogen is required in liquid form. The most efficient way of satisfying this
new demand is to add a liquefier to an existing gaseous nitrogen stream or to
design from scratch a plant with a liquefier. Liquefiers are designed purely
and simply to liquefy a pure gaseous nitrogen stream on a large scale (up to
hundreds of tonnes per day) and can be very efficient. Hydrogen liquefier
plants are usually smaller (10 tonnes per day) but are similarly specialised
and can also be relatively efficient although, as noted in section 2.2.5,
hydrogen requires a large energy input for liquefaction.
Apart from the large liquefiers seen on gas producers sites, however,
small nitrogen and helium liquefiers are often found on-site with their
application. He liquefiers are usually designed as part of a sealed cryogenic
refrigeration package and are usually maintained by the manufacturers.
Small nitrogen liquefiers are often seen on remote sites, however, and are
usually operated to provide liquid nitrogen externally. They are certainly a
convenient
29
source of small quantities of LN, although the economic case
for them is often weak.
Cryogenic insulation. Liquid nitrogen kept in any ordinary vessel quickly
evaporates. Furthermore, the vessel containing it will be swiftly cooled to
near 77 K, at which temperature many vessels crack or are dangerously
brittle. Even worse, liquid oxygen can accumulate on the cold surface,
especially in dry air, and the liquid oxygen can then run off and form a
serious hazard by being absorbed into organic matter. It is not enough, at
least for smaller equipment, to make the insulation thicker. It is not difficult
to show, for example for a cylindrical geometry, that the heat flow Q from a
long tank of radius R with insulation of conductivity S with thickness T is
Hence large increases in T, far beyond R, do not yield very much in the way
of decreased Q and large thicknesses of insulation will not save small
cryogenic tanks from having an undesirably large degree of boil-off. Instead,
the quality of insulation must be improved.
As storage tanks or plant increase in volume their surface area increases
29
They do have inconvenient features, too. Normally a Dewar vessel is wheeled up to these
units for filling. With large vessels precautions should be taken to stop when filling is complete,
as filling is a slow process with these units, unless they have a good size accumulator vessel, and
the operator may be tempted to leave the filling unattended, giving rise to overflows.
only as two thirds of the power of the volume, so that large tanks and plant
can use the crudest insulation.
30
Liquid methane, for example, is often
stored in 10000 ton tanks insulated with simple cement-based insulation
blocks. These would be completely useless for a 251 vessel for liquid helium.
For some applications, such as insulating large cryogenic plant, the use of
good quality mineral insulation, such as perlite, is acceptable. (Even here,
however, there are problems: a nitrogen or dry air purge is necessary to stop
atmospheric air depositing water in the Perlite, for example.)
Ordinary insulating materials are not acceptable, however, in more
stringent applications in cryogenics such as storage or transportation tanks,
or pipes. Most, such as insulation-grade concrete blocks or cork tiles, are
simply not good enough, while others are too expensive or degraded at 77 K.
Very early in the history of cryogenics the fundamentally simple
observation was made that what is not there (a vacuum) cannot conduct heat
and the concept of the Dewar vacuum flask was invented. These are
nowadays familiar as a domestic item. However, they were originally glass
double-walled hand-made glass flasks, blown and then evacuated, made for
laboratory purposes.
31
Vacuum insulation has undergone several improvements since Dewar.
The reduction of radiative heat losses via silvering of the glass was followed
by the discovery that the quality of vacuum needed was considerably
reduced if the vacuum space was filled with a highly porous powder. This is
because of mean free path effects similar to those discussed in section 1.3.13.
Basically, the low-pressure gas in the spaces between the porous powder will
have a low conductivity if the mean free path exceeds the dimensions of the
interstices. The reason for this is easily seen by inspection of the two
formulae for heat conduction per unit area between two plates
At high pressures
While at low pressures
where
P is the absolute pressure, is the differential temperature and L is the
plate spacing.
30
Haldanes famous essay about scale effects in biology (Haldane, 1927) shows the same effect
in reverse: warm animals that live in the cold Arctic without a shelter have to be large in size if
they are to have reasonably low metabolic rates.
31
A standard domestic vacuum flask is still a perfectly serviceable device for storing cryogens.
However, the lid should not be screwed on after fi l l i ng because after a little while the cryogen
will reach a surprisingly high pressure and crack the glass vessel inside, followed by cracking and
leaking of the outer plastic casing. In addition, these flasks will occasionally break on fi l l i ng as a
result of thermal shock.
GAS TECHNOLOGY 227
Note that in the high-pressure case the thermal losses will be independent
of pressure. There is no point in achieving a poor vacuum between the vessel
walls; this low-pressure gas will conduct heat away from the cryogen just as
effectively as atmospheric pressure gas.
At low pressures, when the mean free path becomes large compared to
the insulation cell size, the thermal losses are reduced with pressure, making
the achievement of a better vacuum useful. It is also worth noting that there
is no dependence on spacing L between plates in this case; many plates can
be placed closer together in series and thermal insulation improved without
making the spacing larger. The use of microporous material is thus
indicated.
The cheapest and most popular microporous material is perlite. This is a
natural whitish rock which contains internally trapped water. On firing it
quickly at 800C or so the water boils, creating small pores inside the rock.
The final material is like very lightweight coarse sand. Similar, but coarser
(510mm pieces), is a brownish material called vermiculite. Glass fibre and
similar, but cheaper, rockwool fibrous materials are also useful insulation
materials which can be evacuated to achieve high insulation quality.
The same principles are also exploited in the use of Mylar multilayer
wrapping insulation. This uses multiple wrapping layers of a plastic film,
such as the polyester Mylar or Melinex products, which is then
evacuated. The film has a thin-film deposit of shiny aluminium on its surface
which achieves almost zero emissivity and perfect reflectivity, reducing
radiant heat losses. Evacuated Mylar multilayer insulation is often
dubbed superinsulation and is invariably used for helium vessels and for
the small portable/wheelable vessels often used for cryogens in the
laboratory, which generally lie in the range 202001.
Aerogels. The principle of microporous insulation may eventually be
carried to its ultimate extent in the use of aerogels. These unique materials
are very curious to handle, in appearance something like solidified smoke,
they have densities of the order of a very low refractive index
(1.05) to match, and are transparent with a pale grey/light blue appearance
(Hrubesh, 1987; Fricke, 1988). Their remarkably low refractive index was
the reason for their development. They were first made for their optical
properties, being used in Cerenkov radiation counters in nuclear physics
research. An aerogel produces Cerenkov radiation only with particles
travelling very close to the speed of light.
Aerogels, however, are also extraordinary insulation materials. Instead
of pore sizes in the 0.1 to 1 mm range, these are generally silica-based
optically transparent materials with pore sizes of the order of a few
nanometres (there are also organic aerogels). Aerogels are made by
polymerisation in solution followed by the supercritical extraction of the
solvent. The supercritical extraction is critical, as otherwise the pore
structure would simply collapse from surface tension forces or decompose
during conventional dessication, while the high viscosity of the liquid would
mean that it could not be extracted in any reasonable time.
The currently favoured chemistry for aerogels is to react a silicon
compound, such as tetramethoxysiloxane, with water. The methanol
formed in the reaction is driven off in the supercritical extraction, leaving a
silica nanosponge. An older technology uses silicate precipitated from
sodium silicate (water glass) with hydrochloric acid. The dimensions of the
pores in aerogels are comparable with the mean free path of atmospheric air
and they are being developed to have insulation qualities in the range of 10
times better than conventional plastic foams, without the use of vacuum.
Radiation losses in cryogenic insulation. The losses in cryogenic insulation
due to infrared radiation are considerable. The dependence in the
StefanBoltzmann radiation law means that these losses can be considerably
reduced if multiple layers of low emissivity surfaces can be used; the rather
curious fact emerges that a cryogenic tank can be insulated by means of
layers of metallic foil.
How this arises can be seen if a set of equations describing the heat
transfer q across a set of foils 1 and 2 from (cryogen) to (room
temperature) is considered
When foils are used, there are three linked equations for q
hence, adding the equations,
In general then, adding N foils reduces the IR heat losses by a factor of
N + 1. However, to suppress IR radiation without adding to conductive heat
losses a submicron reflective metal layer on a plastic substrate, typically
vacuum-deposited aluminium, is the most effective technique. Hence
multilayer Mylar superinsulation is always aluminised and flakes of metal
can be added to perlite evacuated powder insulation to minimise IR losses.
Perlite as normally made transmits a certain amount of visible and infrared
light, i.e. it is semi-translucent. Perlite with a little aluminium powder or
flakes added to decrease IR loss is called opacified. Metal flakes can add to
the solids thermal conductivity, however. Perhaps today users of perlite
should consider adding flakes of aluminised glass or Mylar (something like
the glitter used on Christmas cards).
GAS TECHNOLOGY 229
Insulation materials are rarely good mechanically and it is often necessary
to provide better support to the inner wall of tanks. Fibreglass (i.e. glass
reinforced plastic) supports are often used because fibreglass has a
surprisingly low thermal conductivity compared with metals (although the
stainless steel favoured for cryogenic work has one of the lowest thermal
conductivity values amongst metals) (Table 2.13).
Vacuum insulation is often maintained over long periods of time at low
pressure by including a getter material in the annular space of the vessel.
Getter materials are highly reactive alloys which often include finely divided
titanium or calcium (section 2.2.11) and work by chemically combining with
common gas inleaks and outgassing, e.g. nitrogen, oxygen, moisture and
carbon dioxide.
Liquid helium handling. Liquid helium is much more difficult to maintain in
storage than LN or LO. The reason for this lies in its very low specific heat and
low density. For each kilowatt hour of heat absorbed, 221 of LN will be boiled
off or 14001 of liquid helium. Mylar superinsulation and multiple wall vessels,
with liquid helium insulated from an LN outer vessel, are usually employed
for long-term storage of liquid helium or for delivery of liquid helium over
considerable distances. Very large helium tanks will lose considerable
quantities of gas, even in superinsulated vessels, and these vessels are often
fitted with helium recovery. The boil-off gas can be purified and reliquefied
or, alternatively, compressed into cylinders for use as balloon gas.
Cryogenic storage tanks. A typical storage vessel will have a stainless steel
inner vessel and a mild steel outer, with the interspace filled with one of the
Figure 2.57 Schematic of 25 tonne LN storage tank.
cheaper cryogenic vacuum insulations. The interspace is evacuated to well
under 1 mbar, or whatever is appropriate to the insulation, and sealed off.
As well as simply storing the gas, the cryogenic storage tank has to:
be safe
display level or contents
provide an on-line filling facility
supply constant pressure liquid cryogen.

To meet these requirements a simple double-walled vessel must have a
number of other features (see Figure 2.57).
The requirements of safety dictate a number of pressure relief devices.
The interspace must have a non-return valve, which is often simply a plate
held down by the vacuum within on to an O-ring seal. The inner vessel must
be vented by two parallel safety relief devices; typically a bursting disc set at
a few bar less than the safe vessel pressure and a relief valve set at a few bar
less than that. This bursting disc/relief valve set is usually connected via a
two-way selector valve with another similar valve set on the other arm so
that a burst disc or faulty valve can be replaced without blowing 25 tonnes of
product into the atmosphere. These vent valves are connected to the vapour
space above the maximum liquid level. Finally, all sections of pipe
connected to the tank (including the process draw off and vaporiser) which
can be valved off at two ends, must have separate additional safety relief
GAS TECHNOLOGY 231
valves because quite small quantities of trapped LN can rapidly vaporise,
exerting up to several hundred bar and violently bursting pipes and hoses.
The tank contents will usually be read out by a differential pressure
gauge (section 2.3.1) and its operating pressure via another gauge attached
to the same line. It is an advantage to be able to remove faulty gauges with-
out loss of product, so isolating valves are needed.
It is an advantage to be able to fill a vessel that is actually in operation
supplying gas without disturbing the gas-using process by interrupting
supply or changing pressure. This can be achieved by pumping a new cryo-
gen into the tank via two valves: a top-fill valve spraying liquid into the
vapour space and a bottom valve discharging into the liquid. If the deliv-
ered cryogen is cooler than the tank, then top-filling will lower pressure in
the tank and bottom-filling will increase it. In this way disturbance to the
tank pressure can be minimised. Although control via manual valves may
seem crude, it must be remembered that the pressure is governed by the
pressure-raising circuit and by downstream gas regulators in any case.
To fulfil its function properly, liquid cryogen must be supplied to the
process or vaporiser at a constant elevated pressure, typically up to about
15 bar or so. This is achieved by governing the tank pressure via the
pressure-raising circuit and drawing off liquid for the gas-using process via
a dip-tube connection. The pressure-raising circuit takes liquid from the
bottom of the tank, puts it through a pressure regulator and then to a
vaporiser connected to the tank vapour space. If tank pressure falls too
low, liquid cryogen is admitted to the vaporiser and expanded to warm gas.
This sort of pressure-raising circuit has a relatively low pressure feed due
only to the cryogen hydrostatic head (only a few psi if the tank level is low)
so good designs have fairly generous valve and pipe sizing to allow the tank
to be utilised as fully as possible (see Figure 2.58).
Pumps for cryogens. Pumps for cryogenic liquids have some unique fea-
tures which are worth noting. A reciprocating pump is simple enough to
design in principle for liquid nitrogen for gas cylinder filling, for example.
It is simply a reciprocating piston with enough drive to give 250 bar or so on
the outlet and check valves to prevent backflow. However, the check
valves must function in liquid nitrogen, the cylinder must be vacuum-
jacketted to avoid boiling, and the piston cannot be lubricated with oil and
yet must seal reasonably well; leaking past the piston will entail gas loss
and cause heating, leading to boiling. Materials for the cold parts must be
strong at cryogenic temperatures with particular caution over steel and
polymer components. Finally, thermal design must include blocking cool-
ing of the room temperature parts (the electric motor, gearbox, conrod and
their lubricant grease and oil) and allow for rapid contraction of parts as
they cool down.
Commonly, centrifugal liquid pumps are used for low-pressure liquid
transfer around cryogenic production sites and for filling transport con-
tainers. These may rely on thermal blocking from a conventional drive.
However, a neat solution to the problem of a high-speed rotating seal
between motor and cryogen is to enclose the motor in the cryogen; the
so-called canned motor pumps. With many less aggressive cryogens such as
liquid nitrogen and liquid argon (but not liquid helium, where even a small
power dissipation is thermodynamically unacceptable), this is quite useful.
The copper coils in the motor lose much of their resistance at cryogenic
temperatures and run much more efficiently, whilst the heat inleak for the
whole assembly can be lower than the heat input due to resistive Joule
heating of the motor coils (copper resistivity at 77 K is roughly one seventh
of its room temperature value).
Operating cryogen pumps is not as straightforward as operation of
ordinary pumps. During start-up, huge volumes of gas will have to be blown
through the system to cool it down and this must be safely vented. During
cool-down, differential thermal contraction may occur and this may delay
the cool-down. Start-up of the pump itself must be done with caution to
avoid damage whilst the system is cooling down and pumping only gas.
Using liquid cryogen pumps involves some care with design. Most
fundamentally, the suction side of the pump needs to be considered. The
suction side of a pump lowers the pressure on the liquid and, at least during
start-up, the inlet pipework will raise the cryogen temperature. Boiling is
therefore likely, followed, once the pump begins to cool, by cavitation.
32
The increase in speed is accompanied by a decrease in pressure according to
the Bernoulli equation. Cavitation in this context means bubbles formed in
or before the pump which collapse at the outlet of the pump. The bubbles
typically collapse milliseconds after formation. The collapse of a bubble is
accompanied by a sharp pulse of ultrasound which is powerful enough to
erode, over time, even hard metal pump parts and casing. Although some
cavitation is probably inevitable, it should be avoided as much as possible as
it is detrimental to the pump life. The tendency to boil is restrained by the
amount by which the liquid can be superheated but this should not be relied
on. Placing the pump at the lowest point in the system, and using a straight
run of pipe without siphon sections from the supply sump or vessel, is good
practice. Vacuum-insulated pipework is often an advantage in avoiding too
much two-phase flow and heat losses.
The net positive suction head (NPSH) should be calculated for all pump
installations. This is the system pressure (H), plus the tank pressure
minus the cryogen vapour pressure at operational temperature and
32
Cavitation occurs at any point in a system where liquid is subjected to a pressure lower than its
vapour pressure. Cavitation is by no means confined to pump-suction problems. It occurs
around sharp edges in pipework, sudden changes in pipe diameter, elbows, in fact anywhere a
liquid with a high vapour pressure accelerates rapidly.
GAS TECHNOLOGY 233
allowances for pressure fluctuations at the inlet of a reciprocating or
other unsteady pump and flow frictional losses i.e.
NPSH for cryogen pumping is, of course, problematical because of the
large value of With at least equal to in value, H must be large
enough to overwhelm the remaining negative terms. Many pump manufac-
turers will quote mi ni mum NPSH values (often in the region of 1 to 5 m
water gauge) and these can be used for estimating whether or not a
particular installation will function correctly.
Vaporisers. Vaporisers are a form of heat exchanger but one which is not
designed with the usual minimum considerations. In fact vaporisers
operate with a very large but are designed to be as simple as possible and
to have minimum capital cost.
Vaporisers are often simple coils of tubing left in the open for the air to
warm the liquid gas to near ambient. This is fine for systems where the exact
gas temperature does not matter and where sustained high flow rates are not
required. It is a common sight in cold but humid weather, typical west
European conditions, to see massive blocks of icy snow deposits submerging
vaporisers. Ice blocks are often a sign that gas-flow requirements were
underestimated during the installation planning or sometimes a result of
successful company growth leading to gas consumption rates in excess of the
vaporiser capacity. Exceptionally large balls of ice need to be removed
because they can block access to valves and thermally insulate the vaporiser
pipework to some extent, allowing very cold gas through to the point of use.
The effect of the ice ball on thermal efficiency is not perhaps as big as might
be expected, however, because of the effect noted in section 2.4.5;
increasing insulation thicknesses are not very effective.
For higher flow rates it is necessary to add fins to the tubes of these
ambient vaporisers and to operate two banks of vaporisers so that one unit
can be de-icing while the other is in use. A fan and ducting can be added to
enhance the air convection, as flow rates increase, and a hot water de-icing
(deluge) system used to remove the ice build up. If flows exceed the rate at
which even these deluge vaporisers operate, then it is necessary to use
heated water bath vaporisers. These simple but effective devices are capable
of enormous gas outputs. They immerse vaporiser tubing in a bath of water
supplied with site process hot water or in which are also immersed electrical
or steam heating elements.
Siting of cryogenic tanks. Most gas companies have their own rules about
the siting of larger cryogenic storage tanks. For example, for 20200 tonne
liquid oxygen tanks a typical company recommendation would be for the
following safety distances:
Figure 2.58 Pressure raising in cryogenic t ank.
> 3 m: continuous lengths of flammable gas pipelines;
> 8m: pits, drains, flammable gas pipelines including fittings, roads, car
parks, smoking-permitted areas, electrical substations;
> 15m: large wooden buildings, areas where crowds of people congregate,
railways, compressor intakes;
> 30 m or more: bulk flammable gas storage must be placed at even greater
distances, depending on the quantity stored.
Large cryogenic tanks. Unlike the standardised cylindrical shape used for
vacuum-insulated vessels up to a hundred tonnes or so, very large tanks, up
to 10000 tonnes or so, assume very different shapes and designs. As noted
above, non-vacuum insulation becomes more common. Spherical shapes
are also common as the very much higher cost of fabrication can be offset
against savings in material and lower losses.
Cluster tanks are now used where it is necessary to provide a set of tanks
for different cryogenic liquids as, for example, on air-distillation sites where
different grades of liquid nitrogen, oxygen, argon, some at high pressure,
some at low pressure, often require storage. The cluster tank simply groups
all the required storage vessels inside one large cylindrical envelope, the
large interstices between the tanks and the envelope being filled with
non-vacuum perlite or vermiculite.
Cluster tanks also avoid the worst effects of roll-over. Roll-over
manifests itself as an unexpected huge increase in pressure in a large tank. It
GAS TECHNOLOGY 235
occurs when the contents of a tank are stratified, with dense cold liquid lying
above warmer, less dense liquid; a liquid version of the unstable tempera-
ture inversion often seen in the atmosphere. This stratification can be
surprisingly stable until a small disturbance causes some rotation motion to
start. Quickly, the motion builds up momentum, with the cold upper layer
pushing down into the hotter lower layers and vice versa. Suddenly, the
upper layers are hotter, but without the additional hydraulic pressure which
formerly contained them, and they flash off and boil, raising the pressure
suddenly. Roll-overs cause a large amount of gas to be vented and can even
overwhelm safety venting devices and damage an installation.
Large tanks cannot economically be made to withstand much more than
the hydraulic pressures of the liquid sitting in them. They must therefore be
vented to the atmosphere, leading to losses. Losses arise from several
causes:
heating of the cryogen, resulting in vapour boil-off;
breathing of tanks, caused by expansion and contraction of vapour, air

and liquid during temperature cycling.
33
Venting gas can sometimes be employed in some process near to the
storage tank. However, where this is not possible, and where the gas is
valuable, such as helium, then venting losses are dealt with by extraction
from the vent stream and reliquefaction.
2.4.6 Storing gas in gasometers
All over the industrial world large thin-walled steel cylinders with slightly
domed tops, some large enough for a football pitch on the top, used to be a
common sight. These were water-sealed gasometers or gasholders. They are
still a simple and effective means of storing low-pressure gas, and have the
convenience that they accept gas and give it out at a more or less constant
low pressure. They can also be built inexpensively with rather relaxed
standards because of that low pressure.
Mechanically they are simply large versions of the bell-jar in a water
trough system used in chemistry classes at school, save that the bell-jar in this
case actually floats on the water. The pressure from a gasometer is low
(atmospheric pressure plus the pressure necessary to support the bell top
and sidewalls) and rises only slightly with the state of fill as the submerged
walls emerge from the water seal and lose their water buoyancy. To avoid
moisture in the stored gas, other sealing methods can be used: in small
33
The losses due to breathing are not restricted to cryogenic liquids. The typical losses due to
breathing of a solvent tank during filling and emptying, and due to thermal effects, are discussed
in section 3.7.3 with respect to solvent vapour recovery. Solvents in bottles in the laboratory
have surprisingly high losses, ether and acetone are quickly lost, and in hot weather there are
large losses of petrol/gasoline from vehicle fuel tanks.
Figure 2.59 Gas holders using water and rubber seals.
systems, a flexible rubber tube which everts as the gasholder increases in
volume can be used (Figure 2.59)
Today, the storage of methane fuel gas in large bulk is done by
liquefaction of the gas and storage of the liquid methane in a cryogenic tank.
For diurnal storage gasometers are still used, although storage underground
in a wide portion of transmission pipeline (the line-pack) is now the most
common method. During the day, when usage is heavy, the pipeline
pressure is run down, whilst at night it is ramped up. With pressures of up to
75 bar and pipelines of over 1 m in diameter, huge amounts of gas can be
stored diurnally, reducing the pipeline flow losses and decreasing the
required pipe diameters.
2.4.7 Making and using compressed gas cylinders
The earliest commercial supplies of gas in transportable containers were
probably delivered in gas-bags. Gas-bags, despite their bulkiness, are still in
use for small amounts of gas, for example for calibrating analysers or for
transporting samples for analysis. Today gas-bags are made from tough
polymer film laminated with metal, which prevents escape of gas by
diffusion or contamination by air.
Gas can be supplied in disposable containers such as aerosol spray cans.
These can withstand only modest pressures, however (10 bar or so) and are
available only up to about half a litre water capacity, giving contents in
permanent gases of only 51, a somewhat inefficient package. Oxygen for
recreational use is supplied in spray-type cans in Japan. For readily
liquefiable gases such as butane, aerosol cans are more efficient, although
the pressure rise with temperature limits the vapour pressure of the gases
GAS TECHNOLOGY 237
which can be delivered in this way.
34
However, the mainstay of compressed
gas usage is, and will remain, the high-pressure refillable gas cylinder.
Compressed gas cylinders made from strong forged or welded steel or
aluminium, capable of containing 50 bar or more pressure, with reasonable
wall thicknesses have been around since the late 1800s, when the Taunton
and Erhardt processes for deep-drawing thick steel cylinders became
available.
Gas-to-packaging ratio. An early cylinder might have achieved
in other words a cylinder weighing 100 kg would contain only
5 m
3
of gas at atmospheric pressure. Although the materials used now are
stronger, pressures higher and wall thickness somewhat less, it is still the
case that the gas-to-packaging ratio is not very high for compressed gases;
the cylinder far outweighs the product it contains.
This is worst for hydrogen, where a 44 tonne articulated truck can deliver
only about 100kg of Even for the more favourable example of the
oxygen cylinder, the ratio is only about 1:10; the standard cylinder itself
contains less than 20 kg in a 100 kg package. The liquefiable gases such as
carbon dioxide are more efficient, with the efficiency decreasing as the
vapour pressure at 70C (the normal upper temperature limit for safety
purposes) goes up. Gases such as butane in thin aluminium refillable or
throw-away spray-can containers are the most efficient in their gas-to-
packaging ratio.
Unfortunately, there is little that can be done about the problem of
efficiency of gas cylinders for gases which will not liquefy at room
temperature, except the development of stronger container materials. If a
long gas cylinder is considered, for example, it is readily shown that the
weight of the cylinder is proportional to its volume for a given pressure, i.e.
there is little advantage in larger cylinders. In practice, cylinders of below 20
or 301 water capacity are less efficient. Above this size, however, cylinders
are made up to 10001 or more but are no more efficient than the standard
40501 size.
This can be understood by studying the tension in the walls of a gas
container. The tension T per unit length (hoop stress) in a sphere is given by
T = PR/2
34
For larger capacities, plastic containers might be used in the future. With metal or other
coatings to slow loss of gas by diffusion, biaxially-stretched polyethylene terephthalate (PET)
bottles of a few litres capacity could be used. Ratings of 3 barg are readily available (bursting
pressure 12 barg) and probably at least 5 barg is safely possible. These containers would offer an
inexpensive and lightweight throw-away package for up to 151 of permanent gas and much
larger quantities of liquefiable gas. Safety problems with using plastic gas containers may be a
limitation, however.
T = PR/2
where P is the pressure applied and R is the radius of the hoop. This can be
seen most easily if a sphere is imagined to be cut in half along a diameter.
The total tension force across the cut must be and must be equal to the
force due to the pressure inside across the cut area, i.e. The minimum
cross-section of the walls must be proportional to the hoop stress, i.e.
where k is a constant.
therefore
Hence the total mass M of the sphere must be
i.e.
Hence it is seen that the mass of the container is proportional to the
volume enclosed and there is no advantage to using larger containers.
Larger containers, although needing less wall area, must be made with
thicker walls and this latter fact exactly compensates for the lower area
needed. It is also interesting to note that for approximately ideal gases, there
is also no weight advantage in using higher pressures because they simply
need thicker and heavier cylinders. Higher pressures, however, do allow a
smaller volume package.
Gas compressibility (Z-factor) affects these conclusions to some extent,
giving an advantage to higher or lower pressures depending on Z behaviour
with pressure, and liquefiable gases are obviously usually stored in cylinders
strong enough to render them liquid.
The above discussion is centred on spherical vessels but similar consider-
ations apply to cylindrical vessels. However, in cylinders the longitudinal
stress is that of a sphere of radius equal to that of the end caps, but the
circumferential (hoop) stress is doubled, so that cylinders must be made
proportionally thicker. The greater hoop stress also means that cylinders
tend to fail via cracks along their length (like a sausage splitting longitudi-
nal l y in the frying pan).
Fabrication of cylinders. Standard gas cylinders are made by a multistage
deep-draw forging process involving massive rotating machinery and steel
heated until it is bright yellow (Figure 2.60). First, a thick-walled pot is
formed from the heated steel billet in a back-extrusion or piercing process,
then the pot is drawn over a mandrel with powerful concave rollers, which
give the precise cylinder shape and draw out the walls to the correct
thickness. The base of the pot is thinned and dished so that the finished
cylinder will have a concave base which will allow the cylinder to be placed
GAS TECHNOLOGY 239
Figure 2.60 Making a seamless gas cylinder.
free-standing upright. Previous to the latter process, single cylinder
handling in depots and during fillings was rendered considerably more
difficult because cylinders could not be stored upright except in frames or
trollies. Round-bottom cylinders are still favoured for some purposes; when
used, in manifolded cylinder sets, for example, the round bottom is quite
acceptable. Stress concentration occurs in free-standing cylinders which
have to be carefully modelled to ensure that the material is not overstressed,
and the base must be made much thicker than a round-bottom cylinder.
The crucial neck end of the cylinder is then fashioned. Following cutting
to length, the thin-walled pot is then reheated if necessary (by an electrical
resistance heating method in some works) at the neck end and spun on a
lathe. Heavy shaping wheels then squeeze the walls at the end together into
a thickly reinforced neck area. A neck ring can be added at this stage. The
whole cylinder is then heat-treated by furnace heating and oil (or polymer)
quench, cooled to ambient temperature, and ultrasonically tested and
visually examined with a borescope, before being stamped with a unique
number and then shipped. Polymer quench heat-treatment may become a
standard in the future because a polymer/water quenchant mix can be
tailored in the cooling rate it applies, and is intermediate between oil and
water quenching. The cylinder serial number, together with other infor-
mation such as pressure rating and empty weight, are then stamped on the
wider shoulder part of the cylinder neck. A record of each gas cylinder
serial number is kept, so that in principle they are as traceable as a motor
vehicle, with test records also recording the cylinder number, just as for
motor vehicle testing.
Cylinders are made and tested to comply with codes of practice and
regulations which, although similar in effect, differ in numerical details and
procedures around the world. Cylinders in the US must meet ASME or
DOT standards, in the UK they must meet BS standards, and in Europe
TUV, DIN and EC codes. It is essential to refer to these requirements for
details.
Ultrasound testing may be carried out, perhaps magnetic testing, too,
either during or after finishing hot-working. Occasional sample cylinders are
tested to destruction. Hydraulic burst testing, or, occasionally, pneumatic or
hydropneumatic burst testing is carried out, giving the ultimate maximum
pressure of the cylinder. Hydraulic failure tests are conducted behind a stout
wooden screen as failure does not engender much release of energy; a simple
split occurs in the cylinder resulting in a harmless gush of water. Pneumatic
tests may emit fragments and involve a violent explosion, and must be
conducted in a properly constructed bunker with, for example, rows of
15 cm thick railway sleepers to protect the bunker itself from damage.
The ultimate failure pressure of a cylinder will always be considerably
beyond the test pressure, itself normally from 150% to 175%, dependent on
local regulations, of working pressure. Cylinders are designed to fail usually
on a longitudinal line along the body. Cylinders which fail in the neck or base
area are regarded as undesirable. The cylinders cylindrical wall is around 5
to 6mm thick in the case of standard large portable cylinders, whereas
bottom and neck regions are up 12mm or so thick. These destructive tests
(especially the pneumatic tests, which model more accurately behaviour
after failure) also reveal how dangerous a cylinder burst may be. Cylinders
which allow fragments to be emitted like a military bomb are regarded as
unsatisfactory, whilst cylinders which simply split in a ductile manner are
regarded as failing in the best possible way. In this way, in an accident,
although personnel very close to the ductile bursting cylinder might be
injured or killed by the blast or by the gas itself, personnel a somewhat
longer distance away will escape unscathed (see Figure 1.2).
Hydraulic tests may involve plethysmography observing volume
changes of the cylinder while it is pressurised. This is achieved by jacketing
the cylinder in a rigid jacket filled with water, and noting changes in the
water level in a manometer tube attached to the jacket. The plethysmograph
readings can be used to set maximum permissible strains for cylinders with a
simple and consistent apparatus.
Standard gas cylinders have a thread machined inside the neck with a
GAS TECHNOLOGY 241
suitable thread-cutting machine. The valve, which has a tapered thread, is
then screwed, with a suitable thread sealant, into the cylinder to a specified
(very high) torque with another motorised machine. Because of the risk of
neck-crackingifovertightened, or leaks if undertightened, gas cylinder valves
should not be regarded as removable by the user and should always be fitted by
the gas company or cylinder supplier. The cylinder guard, a tough plastic or
metal shield, is fitted over the valve .This provides protection against the valve
being broken off in the event of the cylinder being dropped, while also
providing a grip for more convenient handling and allowing easy access to the
cylinder valve and connection. It is worth avoiding a cylinder valve breakage;
if a cylinder valve is broken under pressure, the broken parts will fly off
dangerously, whilst the cylinder itself, if untethered, may be propelled
(briefly) along the ground, depending on the size of the orifice formed,
causing further damage in addition to that which the gas leak may cause.
An ordinary steel cylinder is often left with its natural finish on the inside:
fairly clean bare steel with a modest amount of mill-scale (black iron oxides
which are formed due to oxidation while the cylinder is hot during the
forming process). The scaling on the outside of the cylinder is typically
rather worse than on the inside and often storage periods before sale of a
cylinder mean that it is often already beginning to develop a thin coat of rust
before delivery. This and mill-scale is removed, usually using shot-blasting,
and one or more coats of paint applied to the outside. Gas companies often
pass their new cylinders through the same shot-blast and paint line that old
cylinders pass through. Cylinders are regularly retested hydraulically and
then repainted; a coloured neck-band indicating last date of retest helps
facilitate this retest procedure. The paint colours follow a code related o the
cylinder contents, which varies from country to country, and is important in
allowing speedy identification of potential hazards in an emergency (see
Appendix D).
Special gases often require an improved finish on the inside. Despite
having to carry out all operations through the neck hole, it is possible to
carry out practically all the normal metal-finishing processes on the inside of
a gas cylinder. Shot-blasting to remove mill-scale, followed by some sort of
plating, is quite common. For stainless steel cylinders, internal electropo-
lishing is often carried out, while aluminium cylinders may be anodised.
There is a large variation in the sizes and shapes of gas cylinders used.
Forged steel cylinders have steadily advanced in pressure over the years,
from 50 to 137, 175, 200 and 230 barg in most applications. The latest
generation of steel gas cylinders to be introduced in the UK are now certified
for filling to 300 barg; other countries around the world are following suit.
The lowest pressures normally possible now are around 125 bar, whilst the
highest pressure single-wall cylinders are 400 bar. Cylinders vary in size from
the 25 mm diameter and 200 mm length of an ethylene tube cylinder, used
for a single dose of ethylene for ripening fruit such as bananas, to the
Figure 2.61 Efficiency of gas cylinders.
immense torpedo cylinders of 700 mm diameter and 10 m length used for
bulk storage of compressed hydrogen. Capacities thus vary from 31 in the
ethylene tube to or so in large hydrogen torpedo cylinders.
Depending on the design and material required, seamless gas cylinders
can be made by methods other than the standard deep-drawing of a pot as
described above. Seamless tubing can be used, by necking down both ends,
for the manufacture of cylinders with valves at both ends. Simple circles of
plate can be cold-drawn into lightweight cylinders.
Stainless steels can be used for containing more corrosive and more
valuable gases, as can Monel and other nickel alloys. Gases such as tungsten
hexafluoride, which are both valuable and corrosive, are invariably
packaged in stainless steel or Monel cylinders. However, these are not
useful for more ordinary gases as the materials are dramatically more
expensive than steel (or aluminium) and have lower strength, and so have to
be made very thick, giving a very poor gas-to-packaging ratio at exception-
ally high cost. The most common material for cylinders is still steel,
sometimes carbon steel, but most often with alloying elements such as nickel
or, especially, chromium/molybdenum or manganese. The highest pressure
cylinders available are now in chromium manganese steel and these achieve
the highest capacity weight ratios, around 0.15/0.18 (or a little higher)
The new 300 bar cylinders achieve a result which is nearly 30% better
still, at least in the largest size. Manganese steel cylinders are still at
around 0.15 or so (Figure 2.61)
For lower pressures seamless cylinders are not essential. Steel cylinders
can be prepared by welding from heavy gauge plate for service up to 100 bar
or more. Similarly, aluminium alloys are used in welded form for low-
pressure liquefied hydrocarbons, chlorine and similar gases. Welded
cylinders are subject to many of the stringent rules to which drawn or forged
cylinders are subject, with additional requirements for X-rays of welds for
GAS TECHNOLOGY 243
higher pressure variants. Welded cylinders have been moving upward in
pressure for some years now, a trend which may be expected to continue.
Aluminium versus steel for gas cylinders. Aluminium has advantages in
small sizes but is less obviously effective in large sizes, for example, because
thick walls are required and fatigue cracking may occur. In larger sizes an Al
cylinder is typically only a little lighter than a steel cylinder with the same
mass of gas. New alloys are needed but these will need to be very thoroughly
investigated before they are accepted. Neck-cracking on some types of
current aluminium cylinders has had the effect of slowing what might
perhaps a few years ago have been perceived as a drift towards Al cylinders
for more general use. (In a threaded gas connection, the greatest stress falls
on the threads closest to the high-pressure side (this is generally where
failure will occur around the stress concentration near the first thread) if no
precautions are used. High-pressure reactor vessels have been designed
using a spring coil insert or a tapered thread to avoid this problem.)
Aluminium is used in lightweight forged cylinders for scuba diving
because their slightly lighter weight is a cost advantage in an application
where even small savings are valuable and it also offers some advantages in
corrosion resistance. There are also occasions when aluminium cylinders are
preferred on chemical grounds as they are less reactive towards most
products (except alkalis) than steel cylinders. In addition, aluminium
cylinders can be made to deliver much lower moisture content than steel
cylinders, even stainless steel. Moisture levels of less than 0.1 ppm can easily
be reached in well-prepared Al cylinders.
Ultracylinders. Some military requirements are for cylinders able to
withstand much greater pressures than are normally supplied in cylinders;
up to 400 or 500 bar, for example, in the gas bottles involved in missile
gyroscopes. There are a number of ways of improving the performance of
the standard gas cylinder and, although none is commercially viable for
general use at the moment, some are used in military and specialist roles.
Cylinders can be wrapped with fibre-reinforced plastic (FRP). If the FRP
chosen is, like carbon fibre, both strong and stiff, then it can take much of the
stress imposed on the cylinder walls. The wrapping can be a simple body
wrapping, which reflects the greatest stress direction of the cylinder and
enables thinner metal walls to be used. Alternatively, the wrapping can be
all-over wrapping, allowing a t hi nner base and neck, too. All FRP gas
containers can be made, but are unusual. They normally require a
gas-impermeable lining, as almost all common structural plastics are
permeable or reactive with gases to some extent, and there is even a danger
that gas could penetrate the plastic matrix and weaken the matrix/fibre
interface or the fibre itself.
Wrapped cylinders were developed most obviously in the 1960s by NASA
to serve its needs for ultralight pressurised containers for fuel and gases in
space vehicles. The stress in a cylindrical pressure vessels sides in the
tangential direction is double that in the longitudinal direction. This is easily
visualised as follows. A cylinder of length L, radius r and thickness has to
withstand a longitudinal tension of from its end caps with a
cross-section of metal of giving a longitudinal stress of Pr/(2t). The
cylinder has to withstand a tangential tension of P2rL in a cross-section of
metal of 2Lt (imagine the cylinder is sliced longitudinally up the middle),
giving a tangential stress of Pr/t, twice the longitudinal stress. The wrapping
of fibre is therefore arranged in a tangential direction to prevent longitudinal
splitting of the cylinder. The pressure rating can thus be increased by a factor
of up to 2.
In practice, this doubling of pressure rating can be achieved with a 10%
increase in mass, owing to the much greater strength/weight ratio of the glass
fibres over the cylinder-grade aluminium. To achieve this wrapped cylinders
are set up to be under a certain amount of prestress tension in the fibreglass
when unfilled. This will ensure that the fibreglass and the aluminium each
take their fair share of the stress when the cylinder is at its maximum
pressure. The prestress tension is achieved by a technique called autofret-
tage, i.e. hydraulically inflating the cylinder to above the yield point of the
aluminium so that the Al flows slightly. On releasing the hydraulic pressure,
the cylinder relaxes back into a state of compression in the Al and tension in
the fibreglass.
Fibreglass-wrapped cylinders are made by first specifying cylinders with
extra-thick bases, then wrapping them evenly with a tow of continuous
glass filaments with a winding machine which applies a constant tension. The
tow is passed through a coating and roller squeezing mechanism prior to
being wound, to ensure that the matrix resin thoroughly impregnates the
fibres. Computer control ensures that the wrapping is done consistently. The
for 24 hours. Autofrettage is carried out to prestress the fibreglass. The
normal inspection and hydraulic test procedures are then carried out. In
order to avoid difficulties with cylinders bursting in fires because of the
glass-fibre failing, the cylinder valve may require a thermal safety vent plug,
perhaps made from a fusible solder material (fusible plugs are also common
on many welded gas containers).
A typical military application for a wrapped cylinder would be for a small
lightweight compressed nitrogen bottle. The nitrogen would be used in the
JouleThomson effect cooler to cool the infrared detector of a heat-seeking
missile, or used to power a microturbine to spin a stabilising gyroscope up to
speed in a second or so. Premium breathing cylinders for divers (and also for
land-borne breathing sets) are now also supplied with glass-fibre composite
wrapping, and there is just a hint that cylinders in some portable applications
may move this way, too, with one or two products being advertised in
green epoxy/glass laminate is then hardened by a gentle bake at 100C or so
GAS TECHNOLOGY 245
lightweight wrapped cylinders. The LAir Liquide company offers a
lightweight oxy-acetylene welding set (Rollerflam) using an FRP-wrapped
cylinder for the oxygen supply. However, wrapped cylinders remain, for the
moment, exceptional.
Spherical compressed gas containers. Something that may surprise the
newcomer to the gas industry is the shape of cylinders. Gas cylinders would
have a better performance (around 30% less weight) if they were they
spherical,
35
so why is it that in the gas industry spherical shapes are rarely
employed?
As noted above, some very large liquid storage facilities are indeed
spherical, as are some helium Dewars internally, but these are most
exceptional. Some gas cylinders were spherical (one example was the
wire-wrapped compressed air container in the VI cruise missile) and
examples of these are still preserved by the science museums in London and
Munich.
There are a variety of practical reasons why spherical gas cylinders are
almost unheard of, however. The gas cylinder valve needs to be installed in a
specially strengthened and thickened part of the cylinder wall, which spoils
the perfection of the spherical symmetry anyway. Spherical containers are
difficult to fabricate, whether by forging and rolling or by welded seam
construction, and, once fabricated, are difficult to store and handle
efficiently. The fabrication of spherical cylinders is clearly difficult to do with
conventional metal-forging and drawing plant, but it is surely to be
anticipated in the future that, using another fabrication technique, spherical
or more nearly spherical cylinders might become attractive.
Cylinder valves and connections. Cylinder valves are required to be
substantially tough, foolproof, long-lasting, corrosion proof, resistant to
heat and extremely leak-tight. A cylinder with a damaged valve, either
leaking or jammed on or off is a potential danger, so standards for valve
quality have to be very high. For service in semiconductors an additional
requirement is demanded that no particulate matter, even a submicron
particle, should be generated in the valve.
Most gas companies now possess (and supply) gas cylinder coffins for
dealing with the occasional problem cylinder of a dangerous gas, typically
caused by a valve failure. These are large, strong, corrosion-proof pressure
vessels with a large access door. A leaking cylinder can be loaded inside and
the access door sealed shut. Even if the cylinder completely discharges
35
The performance would be improved as follows. The hoop stress on the walls of a cylinder of
radius r is, say, T. Suppose a cylinder is 9r long, it will have volume 31/3 and weigh about
A sphere of twice that radius would have a similar hoop stress T and volume of 32/3
but weigh only about . The latter would also have no heavy base, although a similarly
heavy neck would probably be needed.
inside the coffin, the pressure build-up is limited due to the larger volume of
the coffin. It can then be safely transported away for disposal at leisure. A
wet scrubber system can be connected to the coffin, loaded with the
appropriate water, acid, oxidising or alkali wash solution (e.g. sulphuric
acid, sodium hypochlorite or sodium carbonate) and then the coffin
depressurised safely. These are, however, very rarely used. Some gas
companies also supply scrubber systems mounted on wheeled trollies for this
kind of emergency.
Most cylinder valves are of a simple, conservative design, proven over
many years to be reasonably effective. A typical design has a simple steel
spindle, its shaft sealed with a packing gland, sealing on a soft or hard seat in
the brass body. The outlet connections for different types of gases are of
different diameter, thread pitch and direction. Some outlets rely on a
bull-nose-to-cone seating arrangement and these require tightening with a
substantial spanner to achieve a leak-tight fit. Other cylinder connections
employ an polymer ring seal, such as an O-ring or a metal-reinforced seal,
in various ways and this can often be tightened to leak-tightness without a
spanner, although the seal ring must be inspected and replaced from time to
time.
The use of cylinder valves involves a certain amount of common sense.
of gas out) can still be done for the more benign gases, especially where
cylinder valve outlet ports may have collected road dust, rain, etc. The snift
removes the debris, which might otherwise interfere with the seating of the
outlet connection or the pressure regulator downstream. However, easily
flammable gases such as hydrogen, or dangerously reactive or toxic ones
such as silane or chlorine, should not be snifted. If outlet connections are
dirty on these gases, less common now, as many cylinders are supplied with
plastic or metal covers or plugs, then a blast of dry clean compressed air and
a dry lint-free cleaning cloth are the appropriate solution. As mentioned
above, cylinder valves are normally on a tapered thread in the end of the
cylinder but, because they are screwed in with a machine, they are
exceptionally difficult to unscrew. In any case, as remarked above, removal
of cylinder valves should only be carried out by the gas company or cylinder
supplier.
With liquefied gases, dip-tubes may be attached to the valve internally.
Long dip-tubes allow liquid to be withdrawn from the cylinder in an upright
position. Small curved dip-tubes (goose-neck) from the valve to the
cylinder side just below the shoulder allow liquid withdrawal in the
horizontal position. With low rates of withdrawal, cylinders of liquefied
gases such as carbon dioxide simply absorb a small amount of heat from their
surroundings and vapour pressure will be maintained. However, if large
withdrawal rates are needed, then it is better to withdraw liquid via a
vaporiser pipe so that cylinder temperature (and therefore pressure) is
Snifting (briefly turning the valve on and then off again to allow a brief blast
GAS TECHNOLOGY 247
maintained. Liquefied gases normally also require protection against
overheating in case of fire. Either a pressure-operated bursting disc (for
higher vapour pressures, as with liquid carbon dioxide) or a fusible metal
plug (for lower vapour pressures such as propane cylinders) is required and
is usually fitted to the cylinder valve.
For some gases multiport valves are required, typically two ports, one
connected to a dip-tube which extracts liquefied gas, and the other
connected to the vapour space. For the more dangerous gases there is some
advantage in the use of a pneumatically operated valve. This cannot be
turned on by an incautious snifter while the cylinder is in transit but can be
turned on from a safe distance once at its destination. Cylinder valves
incorporating a check-valve function have occasionally been suggested and
would have the advantage that cylinder contamination, by rainwater, oil or
other liquids sucked back after a cylinder is emptied, would be less likely.
However, common cylinder valves do not protect against suck-back, and so
it is useful to include check-valves in cylinder gas manifolds and other
systems.
Finally, it seems likely that the slow adoption of 300 bar cylinders may
take place. If they are found acceptable in every way, they could replace 200
and 230 bar cylinders, making a pressure-reducing cylinder valve useful. So
much equipment in the marketplace is compatible with up to 230 bar but is
not sufficiently rated to be safe at 300 bar, that gas companies are reluctant
to introduce the higher pressure cylinders in case users inadvertently
connect them to old low-pressure equipment and this leads to an accident.
The idea of a small, relatively crude regulator that can be installed within the
cylinder valve to reduce the output down from 300 bar is therefore very
attractive. The new cylinders can then be sold, with all their advantages of
higher capacity per cylinder and better gas/weight ratio, as fully compatible
with existing equipment. Designs from the Ceodeux and Broeder Mueller
companies are now being used in small numbers. They regulate to 60 bar,
well below the maximum rating of all equipment, and give an added bonus
that they will provide a steadier output pressure in applications where a
highly constant pressure is needed during cylinder emptying. They also have
a check-valve function. The only significant disadvantage would appear to
be the loss of a convenient small gauge giving cylinder contents, although
weighscales can obviously be used.
Manifolded cylinders. Multiple cylinders connected by a manifold of
pipework is a convenient way of delivering larger amounts of gas than a
single cylinder can handle. A bundle of K-size cylinders is not inferior in
weight for the same gas volume to a single large cylinder but can be made
into a convenient cuboid shape. Such a bundle is dubbed a manifolded
cylinder pallet (MCP) and may have, say, 12 or 15 K-size cylinders,
containing of gas and weighing 1.5 tons, firmly bolted to a frame on a
Figure 2.62 Theoretical adiabatic temperature rise while cylinder filling.
metal pallet for fork-lift truck handling. The cylinders used in MCPs can be
older round-bottomed types now regarded as unsuitable for individual
service. For hydrogen service large numbers of cylinders are often
manifolded together and bolted to a 30 tonne vehicle trailer. Such a
hydrogen trailer, containing up to gas, can then be driven to the user
and left there.
Also very common is a two-cylinder manifold that can be used to provide
continuity of gas supply. Whilst one cylinder (or MCP) is being changed, the
other cylinder is used to maintain the supply pipeline pressure. Many such
systems are manually operated. However, it is increasingly popular to fit
pneumatically operated valves to the manifolds and to connect these to a
controller box that accepts cylinder pressure transducer or electronic
weigh-scale inputs. The controller can be arranged either to simply change
over from one cylinder to another after pressure falls from the first on-line,
or it can be arranged to switch to a standby cylinder only while the primary
supply pressure is low and to revert to the primary when pressure there
recovers. The latter mode is used, for example, when a single cylinder backs
up supply from a large manifolded cylinder pallet.
Cylinder filling. Cylinder filling cannot in general be carried out simply by
connecting a bulk gas store to a cylinder with a pipe and opening the valve.
Temperature rise during filling is just one of the complicating factors.
Temperatures rise during most cylinder-filling operations (Figure 2.62).
This can be understood in gas filling by considering the first slug of gas that
has been filled to be under a piston composed of the next slug; the first slug
will be compressed and will heat up. If the cylinder were perfectly insulated
and of low thermal mass, the heating would be adiabatic.
The ideal adiabatic temperature rise during filling of an ideal gas can be
GAS TECHNOLOGY 249
roughly calculated as follows. First the rise in temperature T of gas
compressed from to in an insulated cylinder is noted
The initial 1 bar fill of the cylinder will be compressed adiabatically to 200
bar, whilst the next 1 bar slug will be compressed from 2 to 200, and so on,
with the temperature rises then being averaged as the gas in the cylinder
mixes. If perfect mixing is assumed to be taking place during the fill, then the
calculation is fairly simple. A simple calculation assuming the above law and
considering the volumes added from 1 to 200 results in the above
temperature curve, assuming no loss of heat to the cylinder and no mixing
between added slugs of gas; in effect plug flow. Although the peak
temperature and the exact shape of the curve will differ in practice, the
calculation brings out some very pertinent points:
the maximum possible peak temperatures are surprisingly high: this will
not surprise anyone who knows how a diesel engine ignites its fuel by
simple adiabatic compression of air (adiabatic compression can also act as
a source of unintended ignition in oxygen systems; section 2.5.9);
the final temperature with fast filling, room temperature gas can easily be
such as to make the cylinder distinctly hot to touch;
the temperature rise is most marked just after first starting the fill and
declines thereafter.
Of course, when the gas is supplied via a narrow valve restriction, there
may be a Joule-Thomson cooling at that valve but this small amount of
cooling will be mostly negated by heat exchange down a long delivery pipe.
The cryogenic method of permanent gas cylinder filling has the advantage
that it can be arranged to vaporise but not to fully heat the gas, so it can
supply the cylinder at a significantly low temperature, compensating for the
adiabatic heating to some extent.
In filling of liquefiable gases such as propane and carbon dioxide, which
exist in liquid form under cylinder pressure, there is little heating. However,
some precautions are still necessary. Cylinders should be filled from the bulk
liquid supply leaving an ullage space above the liquid. The ullage space
should be sufficiently big that the cylinders can be heated to 70C before
exceeding their pressure rating. However, they can easily be overfilled,
unlike cylinders filled with permanent gases, and then they may burst even if
only heated moderately.
Acetylene cylinders warm up during filling, because of the heat of
absorbtion of the acetylene in the acetone. This is a problem because at
higher temperatures insufficient acetylene will dissolve. Acetylene cylinders
are therefore cooled, typically by water spray, during filling.
The temperature rise during fill of a gas into a liquid, such as acetylene
into acetone, can be roughly estimated from the vapour pressure curve for
the mixture as follows. Assume the vapour pressure P follows the law
where k is Boltzmanns constant, N is Avogadros
number, T is the absolute temperature and is a constant related to the
activation energy for releasing a gram-mole of gas from dissolution. Hence
the thermal energy released on solution is simply where the
volume of gas absorbed (V) and the temperature rise ( T) are calculated
depending on the thermal capacity of the solvent and solid absorbent. For
example, for dissolving carbon dioxide in acetone (acetylene is similar), a
maximum temperature rise of around 30C is anticipated (ignoring the solid
absorber in this example).
Cylinder filling with cryogenic liquid. In this technique, liquid gas is drawn
from a low-pressure storage tank, its hydraulic pressure raised with a
reciprocating liquid pump to well above the maximum 230 bar filling
pressure, and then a metered amount of liquid is admitted to a batch of
cylinders connected on a milking machine style set of flexible connections.
Using liquid filling, 15 cylinders may be filled simultaneously in a few
minutes. Pumping cryogenic liquids up to high pressures is also an efficient
way of filling cylinders because the pump only has to pump a small volume of
liquid; the energy to produce the much larger volume of high-pressure gas
for the cylinder is produced by the vaporisation of the liquid. Attention to
safety is paramount, however, because a cylinder overfilled with liquid gas
will soon rise in pressure beyond safety limits and fail disastrously; the
fracture will be brittle if the steel cylinder has been cooled down by the
cryogen sufficiently before failure.
Gas decant filling of hydrogen. Hydrogen is inefficiently stored in cylinders
but is difficult and expensive to liquefy. Considerable ingenuity has
therefore gone into bulk hydrogen delivery. As an alternative to delivering
cylinders, some gas companies drive a vehicle trailer of manifolded
cylinders, comprising either standard or 20 1 diameter tubes full of
hydrogen, around to a customer. Instead of mechanically off-loading the 20
tonnes or so of cylinders, only the 100 kg or so of gas is off-loaded by a pipe to
cylinders fixed on the users site. However, all the hydrogen cannot be
unloaded. Rather, hydrogen flows from the trailer to the users cylinders
until the pressures in each are the same. This gas decant process does have
some disadvantages, however, in that it is either inefficient in its use of truck
capacity or does not fill the users cylinders to full pressure. It also requires
more cylinders to operate than straightforward cylinder or trailer exchange.
For example, consider the following situations:
Trailer capacity initial pressure 200 bar
User storage
GAS TECHNOLOGY 251
Settled (delivered) pressure ~167 bar
delivered 17% of trailer load
Trailer capacity initial pressure 200 bar
User storage
Settled (delivered) pressure ~133 bar
delivered 33% of trailer load
For large users the most common system is trailer exchange; this is efficient
provided a whole 20 tonne trailer is a reasonable match to demand.
Residual gas: pre-evacuation of cylinders. For lower quality gases it is
sufficient to simply blow down cylinders to close to atmospheric pressure
and then refill them. However, high-quality gases must not be filled on top of
the previous contents of the cylinder. Instead, the gas already in the cylinder
must be blown down to 1 atm, through a scrubber exhaust system if
necessary, and then evacuated with an oil-immersed high-vacuum pump to a
low pressure before refilling.
Carbon dioxide. is one of the few common substances which sublime.
Solid carbon dioxide, dry ice, at 95C evolves gaseous directly
without an intermediate liquid being formed, which is very useful in various
dry ice applications (sections 3.4.2 and 3.7.4).
Standard high-pressure gas cylinders and equipment can be used for
handling liquid and gaseous at room temperature. At cryogenic
temperatures can, with some precautions, be handled as a liquid, just as
liquid nitrogen, argon or oxygen can be. Alternatively, cryogenic carbon
dioxide can be handled as a solid, just like ice, which it resembles to some
extent. There are, however, some variations on the standard equipment
especially for and some special rules for handling the gas. One of the
latter is that in handling cryogenic liquid a certain minimum pressure, 4
bar gauge at least, is always maintained above the liquid. It is quite possible
to vent the gas above refrigerated liquid in such a way that the
remaining boils and then solidifies. This solid can then only be
removed by allowing it to sublime. This limits the possible rate of gas
withdrawal to a very low value and entirely prevents liquid withdrawal.
It is therefore highly undesirable. A possible rescue procedure is to
repressurise the tank containing solid with a fresh supply of gas.
The gas will slowly warm up the frozen solid until it reliquefies but this
procedure could take days or weeks.
The most common storage tanks for bulk are insulated pressure
vessels which store the gas at 5 to 15 bar and, if they are to provide hold times
of more than a week or so, are provided with a small refrigerator system
which can provide enough cooling power to counterbalance any heat
inleaks.
2.4.8 Transport of gases
Delivery by pipe. The most efficient way of moving gas depends on how far
it has to go and how much is needed. For large quantities and short distances
a pipeline makes sense. Industrial gases are rarely used in such large
quantities as methane fuel, and can often be made on site unlike methane
fuel, so pipelines for industrial gases are unusual. However, there are
extensive pipelines for other industrial gases in some regions of the world.
Belgium and northern France, for example, have a network of pipes for
nitrogen and oxygen. Silicon Valley (the area around San Francisco and
San Jose in California renowned for the large number of semiconductor
manufacturers) has a pipeline for nitrogen running to most of the wafer fab
plants there. However, pipelines are limited in their capacity by pressure
and maximum flow rate. Maximum pressures are typically limited to less
than 100 bar; often only 10 or 15 bar. There are further limitations on
pipelines because of the necessity of obtaining wayleaves, i.e. the rights to
dig a hole underneath somebody elses property. The cost of erecting a
pipeline in a desert is completely different from costs in a typical industrial or
city site, which can be measured in hundreds or thousands of dollars per
metre.
Road transport of industrial gases. For the smallest users of gases the most
economical mode of delivery is almost always road transport of cylinders.
Despite the high weight of the gas cylinders, this is most often the best way of
obtaining a dry, pure, gas in quantities of up to a 100 m
3
or so at a reasonable
cost. The high cost of road transport means that much of the cost of the
product boils down to the cost of delivery. A complete cryogenic road tanker
unit (tractor plus specialised tank semi-trailer) will typically today (1996)
cost around a quarter of a million dollars. This must be expensively manned,
fueled and maintained, and these costs can easily exceed production.
From the air-distillation column oxygen may be costing only ten dollars or
so per tonne. The same tonne of oxygen in cylinders is a thousand dollars or
so, a hundred times more expensive. Gas manufacturers trade off the cost of
local production with transport and usually arrive at a compromise where
gas from a handful of ASUs is delivered in liquid form to widely scattered
transported only the last 50 miles or less in that form.
For larger volume users of gases, the best form of gas delivery is mostly
delivery and storage of liquefied gas in cryogenic tanks although on-site
generation of gas is often an alternative. After its liquefaction at a large ASU
compressed sites where the liquid gas is packaged in cylinders and
GAS TECHNOLOGY 253
or liquefier unit, gas is loaded onto trucks which typically are large
articulated vehicles with the maximum permitted road weight and thus
maximum payload. The scaling laws for the cost of building and operating a
road vehicle are such that only a small number of smaller vehicles can be
justified, for delivery to cramped city sites, and most of the fleet must be the
largest types permitted.
Despite the utmost attention to transport costs, it is rarely economic to
transport gases such as nitrogen and oxygen more than 100 miles or so by
road. As a result, most gas manufacturers try to achieve a sprinkling of
production plants this kind of distance apart in their key markets.
The limitations of road tanker capacity can be obviated by resorting to rail
transport, although with some disadvantages. A rail tanker train can be a
valid alternative to a back-up plant because it can deliver gas over hundreds
of miles in quantities of hundreds of tonnes per day.
Liquid hydrogen and helium. Liquid hydrogen and liquid helium are
exceptional with respect to the other gases for a number of reasons. They
both require very large amounts of energy for liquefaction but are also very
inefficiently stored in high-pressure cylinders.
Both gases are delivered in liquid form in some circumstances, despite the
huge costs of liquefaction. Road tankers and ISO containers are used, which
are superinsulated and may also have a liquid nitrogen cooled jacket built
into the insulation. Helium is only commercially extracted in large quantities
in a few places, and therefore is most frequently shipped long distances,
while hydrogen is only rarely required more than a hundred miles or so from
a production site. It is rarely moved in liquid form outside the USA,
although the advent of wafer fab plants requiring the high level of purity
found in liquid hydrogen has changed this picture somewhat. Helium is
frequently required as a cryogen when it clearly must be delivered in liquid
form. Small superinsulated vessels are available with a reasonable hold time
for smaller users.
Gas storage in absorber cylinders. The low density of hydrogen storage in
cylinders is a particularly serious problem for developers of vehicles using
hydrogen engines or fuel cells. Metal alloys which store hydrogen have been
extensively investigated for their potential use in hydride-based rechargea-
ble batteries. There are a number of formulations based on nickel, iron or
magnesium metals which store hydrogen at a sufficiently high density
(comparable to that in compressed cylinders) that they can be considered as
hydrogen gas storage alloys. Relatively lightweight cylinders are used as the
hydrogen equilibrium pressure above the hydride is low. There are a
number of R&D programmes, for example that of Mercedes-Benz and the
hydride specialist firm HWT in Germany, researching the application of
hydride storage systems to fuel-cell and other hydrogen fueled vehicles.
However, although these absorbers offer a conveniently lower pressure
than compressed hydrogen, there are problems. Firstly, the alloys must be
prepared in finely divided form, followed by a sintering process so that both
gas flow and surface area are maximised. Considerable heat is evolved
during absorption, slowing the filling process, whilst on desorption cooling
occurs and heat must be put in to maintain gas flows. Also, on cycling, many
of the hydrides swell and contract with the addition and subtraction of the
gas, and this causes fracturing of the sintered granules, creating problems
with dusting. So far, no commercial hydrogen absorber cylinders are
available.
An absorber cylinder system which is commercially available is the
Nippon Sanso/Matheson SDS (safe delivery source). In this, phosphine or
arsine are absorbed onto zeolite in a gas cylinder at atmospheric pressure.
Zeolites and activated carbon absorbers become saturated, i.e. little more
gas is absorbed for further pressure rises after 10 bar or so, at a rather low
gas/solid ratio (a few per cent is typical). As a result of this, these absorbers
are more useful at low pressures and for modest amounts of gas. The zeolite
in the SDS cylinders increases the amount of gas available from the cylinder
at 1 bara by a factor of six. A filter in the cylinder neck prevents dust from the
zeolite leaving the cylinder in use.
Arsine and phosphine are both normally used at reduced pressure, in the
vacuum of an ion implanter for semiconductor manufacture, so the low
source pressure is not inconvenient, whilst the safety advantages are
considerable since cylinder leaks will not result in these poisonous and
pyrophoric gases escaping. Such escapes, quite apart from their danger to
staff, have an economic implication. If a high-pressure cylinder leaks even a
small amount of arsine, for example, the whole of a wafer fab plant, with
hundreds of workers, would have to be evacuated while further gas escape
was stopped and checks were made before sounding the all clear. One
potential problem with storing such reactive gases in this way is the potential
exothermic reaction of air with the stored gas. However, the fairly low
concentration of the gas in the absorber means that the temperature rise
expected is only 35C, according to the manufacturers.
2.4.9 On-site generation of gas
An alternative to moving gas is to make it where it is needed. If the demand
for nitrogen or oxygen is very large, then an on-site air-distillation plant can
be erected for the gas supply. However, few demands outside large steel and
chemical works are this large. As a result, many decisions about on-site
generation for oxygen or nitrogen revolve around the acceptability or
otherwise of the much smaller and less expensive PSA plant discussed in
section 2.1.6.
GAS TECHNOLOGY 255
The decision to employ a PSA unit will depend on the whole economic
equation of the site where gas is needed and on the purity of gas required.
Economic factors would include:

the cost of electrical power needed

the cost of maintenance

the number of hours per annum gas is required

the peak flow needed

whether any subsequent purifier stage is needed.
In addition, some economic factors will also occur, such as the noise
pollution potential of the plant. Not only will there be a continuous buzz
from the compressor, but also many designs produce a loud rush of escaping
gas on a bed changeover; a cryogenic tank is much quieter.
Basically, a PSA unit will match well to a gas demand that is substantial
(say, greater than continuous, low in purity needs, with no peak
flows and, on a site well away from residences or offices. There is a trade-off
in PSA designs between purity and production rate which favours low purity
(section 2.1.6).
Large-scale methods for the chemical generation of gases are not often
used where physical methods are available (section 2.1.5). However, many
chemical generator methods can be made to work in small sizes, too, and can
in these sizes be competitive with merchant gas. However, many chemical
methods require the delivery and storage of chemicals, which may have
disadvantages similar to those of merchant gas, and all have some purity
problems, for example:
from electrolysers: gas produced contains oxygen and is saturated with

water;
from methanol crackers: methanol needed and a PSA type of plant is

often needed to eliminate CO
x
from output;
acetylene from carbide: purity is unacceptable for welding, for example,

without further processing;
from inert gas combustion generators: purity is very poor.

2.4.10 Gas losses
It is generally accepted that a few per cent of most gases, especially bulk
cryogenic gases, are lost in transit from the production plant to the final end
use. Occasionally, buyers of gas become concerned about this and
sometimes some of the loss can be economically reduced. Some loss
mechanisms have already been discussed; those from large static storage
tanks, for example. There are, however, more possibilities for losses:
36
leaks: underground leaks are difficult to both detect and fix (housekeep-
ing figures, generated by summing equipment gas consumption and
relating it to overall flow rates are often not useful due partly to the
inaccuracy of flow meters and partly because only a fraction of equipment
will typically be fitted with metering);
filling warm cryogenic tanks: release of the vapour is often needed if the
pressure rise is such as to prevent pumping in new liquid;
purging gas: essential to maintain purities after connections or discon-

nections have been carried out;
safety relief valves occasionally dump fairly large quantities of gas;
continuing gas supply to equipment which is out of service: this may be

necessary for some equipment to keep it purged ready for action, for
example;
decanting of pressurised hydrogen (as discussed above): the gas delivery

truck always drives away with a sizeable fraction of the hydrogen delivery,
although this gas will typically not be lost but simply recycled as an
undeliverable inventory of product.
2.4.11 Vacuum pump technology
Vacuum pumps are an important part of industrial gases, both in production
and application. For example PS As can be made advantageously with a
vacuum cycle, the vacuum swing absorption (VSA) process, while purging
of gases is more effective with vacuum. In industrial applications the use of
vacuum pumps is legion; many processes can be carried out in low-pressure
gas, often with considerable gas savings. Low-pressure applications are
often restrained, however, by the relatively high cost (compared to mass
throughput) of vacuum pumps and the additional cost of building contain-
ment which will withstand air pressure at 1 bara.
To appreciate the cost of vacuum, consider that a large vacuum pump,
costing 20 000 or more, might typically be pumping a 10 mbar load at
per hour. At this pumping rate, this large machine is only pumping a
kilogram or so of air each hour. To appreciate the cost of vacuum
containment vessels, imagine a 1 kg bag of sugar balanced on every square
centimetre of a large piece of equipment. It is not difficult to understand why
36
The author has on several occasions gone to review gas consumption on complex sites where
there are many different uses for the gas because there has been a suspicion that gas usage was
higher t han necessary. In most of these reviews some unnecessary losses were quickly found but
nevertheless the overwhelming majority of the losses found could not be reduced economically.
There were even one or two pieces of equipment where gas flow had been turned down too low
for safe operation or was reducing equipment lifetime, and an increase in consumption was
recommended.
GAS TECHNOLOGY 257
much stronger and more expensive construction is needed. A warning might
be added here that improving a process by adding a vacuum cycle needs
careful calculation on construction strength; simply applying a modest
vacuum to equipment not designed for it may cause buckling.
The relatively low cost of inert gases such as nitrogen, or even argon,
means that they compete strongly with vacuum as a means of avoiding
oxidation of sensitive or hot oxidisable materials. The use of inert gases at
atmospheric pressure allows vessels of the thinnest metal sheeting to be used
and also allows relatively imperfect sealing to be adequate, with careful
attention to the leaks created. Inert gas would therefore be the correct
choice for a belt-furnace. The vacuum seals needed for a vacuum-belt-
furnace would be almost impossible to engineer for practical cases.
Distinction should be made in vacuum pumps between those capable of
evacuating to atmosphere and those, such as diffusion pumps and turbo-
molecular pumps, which only function with another backing pump to boost
their output pressure up to 1 bara.
Venturi, ejector and diffusion pumps. One of the simplest means of
generating a low pressure is to use a high-pressure jet. A jet of gas or water
can entrain another gas stream or the low pressure caused by a venturi can be
used to suck another gas stream. One of the simplest embodiments of the
entrainment principle is the water-jet pump commonly used in laboratories
for vacuum-aided filtration funnels.
A water-jet pump connected to a rapidly flowing stream of cool tap water
can generate less than 50 mbar at flow rates of a few litres per minute; a litre
of gas is pumped for every three or four litres of water passed. Water-jet
pumps rely on entraining and then lowering the pressure of the pumped gas.
Ejector or venturi pumps which use gas instead of water, are very similar and
exploit the entrainment and venturi principles in a straightforward way.
Diffusion pumps (Figure 2.63) carry the entrainment principle to
ultrahigh vacuum pressures. Although less popular nowadays, these are
very simple, reliable pumps with no moving parts. Once pressure is below a
few millibars or so (where diffusion pumps work best) gas molecules do not
undergo collisions very often, at least over a short distance. It is therefore
possible to consider the diffusion pump as a jet of heavy molecules simply
knocking the light gas molecules out of the pipe being exhausted. A
diffusion pump has a cascade of such jets fed with a copious supply of heavy
molecules arranged in concentric fashion, the higher jets pumping gas to the
next set of jets and so on. The jets further down the cascade become smaller,
more numerous and closer to the pump wall, so that they remain on a scale
comparable with the mean free path of the gas as the gas pressure is built up.
The assembly is capable of pumping down to mbar or so from 1 mbar;
an impressive compression ratio. The jets are fed from the boiler unit and
filled with oils that are resistant to decomposition at their boiling point. The
Figure 2.63 Diffusion pump (courtesy of Edwards High Vacuum, Crawley, UK).
oil boils and the vapour is fed to the jets. Mercury was formerly used but has
a higher vapour pressure than vacuum oils and is now regarded as a
significant health risk.
Mechanical vacuum pumps. The standard mechanical vacuum pump is
still, after a century or so, the oil-immersed rotary vane pump (Figure
2.64a). This is robust and, thanks to being bathed in oil, fairly long lived. A
central rotor rotates in an eccentric housing, whilst spring-loaded vanes or
GAS TECHNOLOGY 259
flaps seal the annular gap into compression chambers. An inlet check valve is
usually needed to prevent oil being sucked back along the inlet line when the
pump stops with its inlet line under vacuum. Two stages driven on the same
shaft are normal, achieving spectacular compression ratios of 10 000 000 or so.
One company (Edwards High Vacuum) is now marketing two innovations
in the primary pump market. One of these is a dual-stage oil-immersed
pump which has two switchable modes to optimise either its mass-flow
throughput or the ultimate vacuum achieved. The other innovation is a
revisitation of the scroll pump principle. This avoids some of the
oil-immersed vane pump problems while being usable in a much smaller size
than the Northey claw pump described below.
For large pumps, the oil-immersed pump is challenged by the Northey
interlocking claw concept, the so-called dry pump (Figure 2.64b). Particu-
larly in the semiconductor industry, where pumps are required to keep the
vacuum process chamber absolutely clean, and where some of the gases
pumped can react with the inside of the pump and its oil, these have
important advantages. Three stages, with the first stage often a high-
throughput Rootes blower, all driven from the same shaft, is a typical
arrangement. Because the rotor tips and ends are not sealed with oil, these
pumps are made with very fine clearances and must be carefully temperature
controlled to avoid problems with clashing rotors. They are the most
expensive type of vacuum pump.
The type that is rapidly becoming the standard high vacuum and ultrahigh
vacuum pump is the turbomolecular pump (Figure 2.64c). This is a
straightforward turbine pump normally arranged with multiple rotor fan
wheels on a single shaft interleaved with stator fans. At the input
(high-vacuum) side, the fans have large, wide, high incidence angle blades.
The wheel and stator fan blades then progressively become narrower and of
lower incidence moving down the shaft towards the outlet, reflecting the fact
that they must pump denser gas.
Turbomolecular pumps necessarily rotate at high speed with small
clearances between rotor blades and the body, and this is an arduous service
for conventional ball and roller bearings. This has led to the adoption of
magnetic bearings in many units (section 2.1.3). Turbomolecular pumps of
standard design cannot pump gas up to pressures beyond a few millibar and
are therefore only operated with a rotary or Northey pump to exhaust their
throughput at atmospheric pressure.
There are two other kinds of vacuum pump, ion pumps and cryopumps,
which are mostly seen in very low-pressure work and therefore of less
interest to us here. The former function by trapping gases into a sputtered
layer of metal in a plasma discharge electrode unit. The latter simply use
liquid helium temperatures to freeze out gases onto simple baffles or into a
liquid nitrogen or liquid helium cooled carbon absorber (section 2.5.3).
Operational pressure ranges for vacuum pumps are given in Table 2.14.
GAS TECHNOLOGY 261
Figure 2.64 Mechanical vacuum pumps: (a) rotary vane pump (courtesy of Edwards High
Vacuum); (b) multistage claw (Northey) vacuum pump (courtesy of Edwards High Vacuum);
(c) turbomolecular vacuum pump: 1, high vacuum connection; 2, rotor disk; 3, stator disk; 4,
venting connection; 5, molecular drag rotors; 6, molecular drag stators; 7, fore vacuum
connection; 8, motor; 9, high-precision ball-bearing with ceramic balls; 10, operating medium
reservoir (courtesy of Balzers Pfeiffer).
Gas ballast. Gases at low pressure can consist partly or entirely of vapours
which will condense at room temperature and pressure to liquids. If this
happens inside a mechanical vacuum pump it can be a problem because the
liquid formed will mix with the pump oil, reducing its effectiveness and
carrying vapour back to the inlet. This problem can be reduced by running
the pump warm or hot; up to around 100C is easily achieved. However, the
use of a gas ballast is often more effective. The gas ballast, typically a small
flow of nitrogen, can be added either to the second stage of a multistage
pump or to an auxiliary inlet port in a single-stage pump. The effect of the
added gas flow is to increase the temperature of the pump, which helps
indirectly, and to increase the proportion of non-condensable gas in the
pump outlet when it exhausts to atmosphere.
To illustrate how gas ballast works, consider a pumping workload
consisting entirely of water vapour. This cannot be exhausted to atmosphere
unless the pump is run at over 100C. At high throughputs, the pump oil
would soon be reduced to an opaque emulsion with a high vapour pressure,
reducing its ultimate pressure. Contrast the situation with nitrogen ballast
added at 100 times the inlet water vapour rate. The pump can now run at a
safer 30C and exhaust to atmosphere a nitrogen gas with 10 mbar water
vapour, since the saturated vapour pressure of water is only 42 mbar.
2.5 Safety and environmental issues
There is a justifiably greater emphasis on safety today than existed even 10 or
20 years ago.
37
The larger gas suppliers all have technical safety departments
who will be pleased to help any customer with safety issues, often free of
charge. If there is in any doubt whatsoever about safety in a gas installation
then professional advice of this sort should be sought.
In many cases, before the gas suppliers will connect up large supplies of
gas, or dangerous or unusual gases, they will frequently insist on qualifying
the user and his equipment as suitable. Occasionally they will perhaps carry
this as far as a safety audit or HAZOP: two species of precise analysis of the
failure modes, their probabilities and consequences for a plant or piece of
equipment.
2.5.1 Where does gas go in an accident?
The most frequent cause of a gas accident is gas escaping from its
containment, either into another part of a process or to the atmosphere.
37
Who today would inhale a lungful of heating gas and proceed to blow it down a Bunsen
burner with the burner lit up? Yet this sort of lunatic party trick, which could easily kill the
perpetrator, was regularly tried by schoolboys in England earlier in the century. Even official
work was not so conscious of safety issues in the past. A common lecture demonstration was the
fi l l i ng of a heavy glass gas-jar with a stoichiometric mix of acetylene and oxygen. The audience
would be rendered temporarily deaf by an ear-splitting detonation, whilst the gas-jar, inside a
length of iron pipe, was reduced to a heap of powdered glass. A remote concrete bunker
equipped with a robot to ignite the gas would probably be de rigueur today.
GAS TECHNOLOGY 263
Where the gas goes in the event of a leak or equipment failure is often the
most fundamental question in a hazard analysis. Is the gas lighter or heavier
than air? Is there a pit in the floor or a trap in the ceiling where accumulation
could take place? Is there a source of ignition in the place of accumulation?
If an operator were rendered unconscious by the gas, would he fall into a
zone of more danger or a zone of less danger?
That gas does sometimes get into very peculiar places is not in doubt.
Much has now been written about the percolation of radon gas from
uranium in some granitic areas such as Cornwall in England. Huge increases
in lung cancer were related to radon inhaled by the occupants of houses with
stone floors. Less well known (Hendrick and Sizor, 1992) are incidents due
to old mine workings, some hundreds of years old. Mine workings undergo
breathing, i.e. exhaling the carbon dioxide- and oxygen-depleted air
contained in them, during excursions in atmospheric pressure. In some areas
of northern England this gas, picturesquely known as stythe, is seen in
houses above the workings even years after closures, and can be dangerous.
Nitrogen gas from cryogenic leaks flows downward; although is lighter
than air at the same temperature, the low temperature of means that it
will be denser than air. This has led to fatal asphyxia accidents when
attempts have been made to fix cryogenic leaks occurring in pits or in
enclosed buildings. Strong ventilation, i.e. opening doors and windows and
if possible a powerful fan, is needed to clear such gas accumulations by
blowing in fresh air if an enclosed or pit area must be entered.
Finally, along with elaborate precautions to ensure safety with gases
should go a reasonable degree of prudence with respect to the more ordinary
accidents it is possible to have with gas equipment. Probably the most
common accident in the gas industry needs no more than an empty cylinder
and a little ordinary carelessness: the toppled gas cylinder. A gas cylinder is
sufficiently heavy (up to 100kg or more, depending on the exact size and
contents) to be capable of spraining ankles, breaking toes and causing
unpleasant flesh wounds, irrespective of its contents. Gas cylinders should
be secured to prevent toppling when not actually being moved.
2.5.2 Toxic and asphyxiant gases
When asked about gases, heating gases usually come to mind first. Second
thoughts might revolve around the poisonous nature of some gases.
38
It is an
unfortunate fact that very few gas mixtures are safe to breathe. Even gases
which are not actually toxic may have other undesirable consequences and
38
My grandfather saw chlorine gas used on the Western Front in Belgium in the First World
War and talk of gases would quickly remind him of this, whilst millions of not-so-old people
remember the gas masks that were carried, but fortunately never used, in the Second World
War.
all gas mixtures containing insufficient oxygen will cause asphyxia. Some
examples of typical classifications of gases and their toxic effects, giving the
USA National Institute for Occupational Safety and Health TLV (threshold
limit value) or IDLH (immediately dangerous to life and health value) are
given in Table 2.14. The TLV is the maximum concentration permitted
during the working day and is similar to the OES (occupational exposure
limit) for 8 h TWA (time weighted average). As can be seen, descriptions do
not follow a totally consistent pattern. Butane is probably a more powerful
anaesthetic than for example, whilst highly toxic chlorine has a higher
IDLH than merely toxic ozone. Some of these discrepancies reflect
general practice. For example, ozone is rarely used in concentrations higher
than a few per cent, so leaks in practice are less serious than leaks of 100%
There are large numbers of gases that are so dangerously reactive or toxic
that most countries insist on special licenses for businesses that use them.
The military poison gases (actually, most of them are liquids with an
appreciable vapour pressure at room temperature) are some of these. After
a initial crude attack in 1914 with cylinders of liquid chlorine, the deadly
business of war moved on to expressly designed chemical poison molecules
such as phosgene Lewisite and mustard gas. These fairly simple
compounds do have some peaceful uses. Phosgene is a relatively common
chemical intermediate and other materials of this type are occasionally
required in small quantities for scientific research purposes. Mustard gas, for
example, can be used to induce mutations in cell genetic material.
The organophosphorus poison nerve gas compounds Sarin and Tabun
are much more complex molecules that work by interrupting normal
GAS TECHNOLOGY 265
biological cell function. They inhibit the enzyme cholineresterase used in
biological nerve systems and are extremely dangerous. They are related to
the organophosphorus insecticides developed in the 1930s and almost
unused in war, save for some incidents in the Iran/Iraq war and a recent
terrorist attack in Tokyo. Their ability to interrupt particular biological
reaction sequences means that there may be occasions in which milligram
amounts are needed in biochemical research.
There is general agreement worldwide about toxicity of gases. However,
in each country there are separate and different regulations about TLVs and
IDLHs, and levels vary somewhat, as do the descriptions applied to gases.
Double-pipe containment. Double-pipe containment is occasionally used
for highly toxic gases. Longer runs of pipe, usually those in between the
source cylinders and the equipments using the gas, are fabricated in two
concentric pipes. The normal gas path is down the inner pipe. The outer pipe
is purged continously with nitrogen and vented to an exhaust stack or to a gas
scrubber. Ideally it will be monitored for toxic content.
2.5.3 Cryogenics
Safety with cryogenics is mostly a matter of common sense. If cryogenic
surfaces are treated with the respect accorded to hot surfaces, then few
problems will in general be encountered. Cool surfaces should not be
touched except with gloved hands and contact with cryogenic liquids should
be avoided.
Some additional phenomena need to be allowed for, however, relative to
hot liquid handling; one is the welding of human flesh to cold metal and
another is the brittleness of many materials at low temperatures. Most
people have touched the walls of a domestic freezer and are none the worst
for their experience. Similarly, most of the time, the metal components of a
cryogenic liquid system, even those at very low temperature, can be touched
briefly without apparent harm. This is because cold surfaces in a typical
ambient atmosphere quickly condense moisture, accumulating a protective
film of ice. However, just occasionally, the metal touched will be relatively
clear of the condensed ice that quickly forms on a cold surface, and this clear
surface is very dangerous. The welding of flesh to the cold metal is instant
and often the cause of unpleasant wounds.
There is a temptation, when a moving component is rendered immobile,
frozen, by low temperature, to hit it with a hammer. This must, however,
be strenuously resisted because normally robust steel or plastic parts may
shatter and the small problem of sticking caused by the low temperature may
become a much bigger problem of broken components.
Leidenfrost phenomenon. There is an effect, the Leidenfrost phenom-
enon, which can give a degree of safety in handling cryogenic liquids. The
occurrence of the Leidenfrost phenomenon is not confined to cryogenics and
in fact occurs in many places in physics. Leidenfrosting is seen on scales from
the largest to the smallest; from the collision of galaxies made of antimatter
and matter, to the hovering of droplets of water on a hot saucepan.
Basically, a cryogen such as liquid nitrogen (LN) dropped onto a much
warmer surface forms and maintains a thin film of boiling vapour which
isolates the LN from the surface. In this way, personnel are to some extent
protected from small accidental spills and droplets landing on them from
boiling LN because the droplets do not wet skin easily because of
Leidenfrosting and tend to run off harmlessly, even more effectively than
water from a waxy surface.
This can lead to a false sense of security, however. If the LN pools in the
surface that it lands on or otherwise remains there long enough, it will cool
that surface until the rate of vapour production slows. At some point the
protective film of vapour disappears, allowing a suddenly much increased
rate of cooling and resulting in painful and sometimes dangerous burns to
personnel. Finally, it is worth mentioning that surfaces cooled in dry air by
liquid nitrogen can condense liquid oxygen, which is itself a relatively
dangerous substance (section 2.5.9).
2.5.4 Flammability
Many gases are inflammable and all flammable gases have the potential to
cause explosions. Some gases ignite or explode spontaneously, such as
silane. Gases specifically designed to ignite on contact with air or oxygen are
sometimes dubbed hypergolic. Compounds such as pentaborane or triethyl
aluminium reliably ignite and have been used in small rocket motors (silane
ignition is not reliable, particularly for pure grades).
Flammability is a very complex subject in general but has been intensively
studied over the years. For combustion, certain conditions must be met
which are characterised by more or less quantitative parameters:
the fuel/oxidant ratio must be correct within limits: these are the upper
and lower explosive limits (UEL and LEL);
the oxidant stream must contain a minimum amount of oxygen: this

percentage of oxygen is called the oxygen index;
a minimum amount of energy must be provided to the combustion

mixture: this is the ignition energy;
GAS TECHNOLOGY 267
a minimum temperature of ignition source must be applied: this is the

ignition temperature.
Flammability of mixtures in air at atmospheric pressure, at least in safety
matters, is the most common situation so this is mostly what is considered
here. Flammability can be considerably enhanced by oxygen enrichment
(section 2.5.9).
39
Flames. It is possible to react together fuels and oxidants using a
heterogeneous catalyst; 1% platinum on a ceramic fibre plate will allow the
combustion of common fuels from around 300 or 400C upwards. However,
a catalytic system is not necessary because of the phenomenon of stable
combustion flames.
A flame is the reaction zone between a fuel and an oxidant gas (Figure
2.65). The zone typically appears more or less stationary but is in fact a
dynamic phenomenon formed out of the fast flowing gas mixtures. A
conventional flame burns at a rate which equals the flow rate of the gases so it
is in dynamic equilibrium. A flame is a self-sustaining chain reaction: once
started, the flame generates heat, which raises the temperature of more fuel
and oxidant so that it, too, combusts, generating more heat, which raises the
temperature of more fuel and oxidant, and so on.
The most useful distinction that can be made is between diffusion flames
and premixed flames. In the former case, pure fuel flows into the
atmosphere, and the flame is formed where fuel and air mix a large,
typically luminous, flame. In the latter case, the gases are premixed and
allowed to flow from their containing tube into the atmosphere, where they
are ignited. The fuel and oxidant are raised in temperature above the
ignition temperature for the mixture as they approach the flame and the
reaction zone begins. A complex molecular dance then occurs, with
molecules losing and then gaining electrons and atoms, and the temperature
rises rapidly, limited only by reaction kinetics to reach typically 1800C or
more. Within the space of a millimetre or so the luminous part of the flame
has begun. Finally, reactions proceed to reach equilibrium and the final
temperature (the adiabatic temperature if the flame is not cooled) is
reached in the bulk of the flame.
A majority of relatively stable species, such as carbon dioxide and water
vapour, are then formed as ambient gas entrainment and radiation cools the
gases beyond the tip of the flame to below 1000C. Mixed in with stable
39
Similarly, flammability is affected by pressure; the Victorian fire piston was a demonstration
showing just that. In the fire piston a piece of highly flammable solid material at the bottom of a
cylinder, which can be made of glass, is ignited simply by compressing the air above it suddenly
with a piston. This both raises the temperature adiabatically (see section 2.5.9 adiabatic ignition
of oxygen mixtures) and raises the oxygen partial pressure, which itself renders the mixture
more flammable. Fire pistons were one kind of portable convenient source of ignition before
gaseous or petro fuelled lighters lit by misch-metal flint or piezoelectric spark became
available.
Figure 2.65 Simplified flame structure.
species will be low levels of undesirable gases, NO and (collectively
called if air has been the oxidant, along with partially combusted fuel,
sulphur dioxide from any sulphur in the fuel and, with hydrocarbon fuels,
microscopic carbon particles. The heat released sustains the chain reaction
and tends to make it propagate back down the supplying tube at a fairly
well-defined velocity known as the flame speed. If the flow speed is lower
than the flame speed, the flame will travel back up the tube; this is called
blow-back. On the other hand, if the flow velocity is too large, the chain
reaction cannot propagate back against the flow fast enough and blow-off
or lift-off of the flame occurs. Stabilisation of the flame is achieved in small,
simple burners by the cooling and flow slowing effects of the edge of the
supplying tube. In larger burners specific measures such as a quarl (a small
ante-chamber into which a furnace burner fires), rapid rotation of the gas
stream or a bluff object in the gas stream, are used. These cause flame
recirculation, giving stabilisation.
Diffusion flames are only rarely used. Diffusion flames are typically large,
luminous, unstable and, if using a carbon-containing gas, deposit carbon
onto cool solid surfaces placed in them. Premixed flames are those normally
employed in most applications. They are typically much less luminous than
diffusion flames, are stable and are much smaller than diffusion flames with
the same thermal output. Partially premixed flames are very common and
they have two often distinct flame zones: a small inner zone of reaction of the
premixture and a large, less luminous outer zone where oxidant diffusing
into the still combustible reaction products of the inner flame allows further
combustion.
GAS TECHNOLOGY 269
There is a further distinction amongst flame types depending on their flow
characteristic. While small burners (such as cooking burners or laboratory
Bunsen burners) tend to have laminar flow, larger industrial burners show
turbulent flow. In turbulent flames, the reaction zones described occur in
small transient combustion cells formed in the intensively mixed gas.
Turbulent flames are described as premixed or jet flames. In the latter,
although the fuel supply is separate from the oxidant, the turbulent mixing
means that, in effect, they have characteristics very like premix flames.
Turbulent flames can achieve greater combustion intensity (throughput of
reactants per unit volume of flame) than laminar flames and many industrial
designs rely on this to achieve high rates of heating and effective heating at
high temperature. High intensity flames are unstable and have a strong
tendency to lift-off. One stabilisation system for intense flames is an oxygen
pilot. A small side stream of oxygen is used and the oxygen locally increases
the flame speed, allowing a more intense flame before lift-off occurs.
Ultrahigh intensity burners sometimes use a combination of laminar or
low-velocity flames with high-velocity highly turbulent flames. The lower
velocity flames stabilise the unstable high-velocity flames, relighting them
continuously.
To achieve a premix flame from a liquefied gas source involves vaporising
the liquid stream effectively. Simple burners rely on the burner parts heating
up to do this. Other burners use an ambient air-heated vaporiser. Hot-air
balloons have special vaporiser coils near the lower part of the flame to
ensure that the liquid propane fuel used is thoroughly evaporated into the
gas prior to combustion (see Figure 3.45).
Rocket motors have a similar arrangement whereby liquid oxygen, and
liquid hydrogen if used, is fed into pipes welded onto the outside of the thin
metal propulsion nozzle, ensuring complete conversion into gas using the
formidable heat of the rocket flame. The latent heat of vaporisation of the
cryogens cools the rocket motor walls, allowing them to survive the intense
conditions for longer.
2.5.5 Flammability limits
Most substances that are not fully oxidised and therefore inert will display
a positive heat of combustion, according to standard chemical enthalpy
tables. However, compounds such as perfluorinated alkanes have very low
heats of combustion (less than 8000 and will not normally ignite. Other
compounds may be so diluted with inert materials, as with the asbestos
/rubber composition gaskets popular in gas systems, that their net heat of
combustion is low. With these exceptions, all compounds are flammable and
solids such as polymers PE or nylon (combustion heats 42 000 and
34 000 approximately, respectively) will burn vigorously. Solids burn
by reaction at their surface with the air or oxygen in the same way as a
candle, the tendency to ignite being formalised in the oxygen index
measurement. Liquids and gases can also be ignited in this way and burn at
the surface, giving a diffusion flame. However, in the case of vaporisable
liquids or gases they can mix with air or oxygen and combustion can take
place within the bulk of the mixture, often explosively. The tendency to burn
in bulk in this way can be given a quantitative shape in the form of
flammability limits.
If a gas is mixed with air in stoichiometric proportions, i.e. the oxygen in
the mix is exactly that needed to oxidise the fuel, then, if the gas is at all
flammable, it will ignite. For instance, a mixture of 29% hydrogen to 71% air
ignites explosively. However, if the gas is mixed in proportions far from this
ideal mix, it may also ignite explosively. The limits of percentage in air which
ignite are an important safety parameter. The lower explosive limit (weakest
fuel mixture) is important in situations where fuel gas leaks from a pipeline
or vessel into the air, where there may be a source of ignition. It is used in
setting flammable gas detectors; these are frequently set to alarm at 20 or
40% of the lower limit. The upper explosive limit is important in situations
where there may be an ignition source within a fuel stream or vessel.
Furnaces for steel heat treatment, for example, involve temperatures well in
excess of ignition point with a fuel gas atmosphere. In the event of
incomplete purging of air from the system, for example during start-up, the
proportion of fuel in the mix may sink below the upper explosive limit and
become dangerous.
Some gas flammability limits are listed in Table 2.16. Outstanding
amongst them are the extraordinarily wide limits of hydrogen and acetylene.
The flammable limits of mixtures are not in general predictable and should
be measured in cases where they are important. However, early in the
investigation of combustion processes in France, le Chatelier proposed an
approximate formula on the following lines. A mixture is flammable if
where are the mole fractions in the atmosphere of flammable
fuels 1, 2 , . . . and so on. This common sense and simple formula can,
however, be quite inaccurate in particular cases.
GAS TECHNOLOGY 271
2.5.6 Ignition energy
If a gas mixture is subjected to a high-temperature spark, it may not ignite.
Anyone who has snuffed out a lighted match in a saucer of paraffin (an
experiment to be tried only outside with small quantities) will have an idea of
what is going on here. That same saucer ofparaffin will be easily set alight by
a roaring gas burner, whose flame may well be cooler than the lighted match
but whose energy is larger.
For most gas mixtures flames only occur if a certain minimum amount of
energy is supplied: the ignition energy. For the hypergolic fuels, the ignition
energy is effectively zero; a passing cosmic ray or a random conjunction of
molecules is sufficient to start the chain reaction. Examples of hypergolic
fuel mixtures are impure silane and air or unsymmetrical dimethyl hydrazine
and dinitrogen tetroxide. The latter mixture is frequently used in aerospace
applications where cryogenic tanks are not wanted.
For some gases, such as hydrogen, the ignition energy is exceptionally
low, just a few microjoules. This range of ignition energy can easily be
generated in static electrical discharges which occur in everyday activity.
40
For other fuel gases, ignition energies are in the millijoule range but still
smaller than the static voltages which can occur. Proper precautions to avoid
all high voltages are necessary to avoid accidental ignition of gases where it is
necessary to use apparatus in an inflammable atmosphere (section 2.5.8).
The minimum ignition energy is usually estimated by gradually increasing
the size of an electrical spark in a stoichiometric mixture of fuel and oxidant.
The spark energy is related to the energy stored in a capacitor discharged
across the spark gap.
2.5.7 Oxygen index
The oxygen index (OI) of a substance is the lowest percentage of oxygen in
the air which still allows self-sustaining combustion of that substance. It is
determined by igniting a sample strip in a steady stream of mixture
and seeing whether the sample stays lit like a candle or extinguishes. Like
the LEL and UEL, the OI is a fairly quantitative concept, although OI is
dependent on pressure and is affected by a number of other considerations
as well as ignition energy and temperature. It is interesting to note that many
polymers, even those such as PTFE which are generally reckoned to be inert
and nonflammable, have an oxygen index, although only at elevated
pressure in the case of PTFE (Table 2.17).
40
A plastic biro pen rubbed on a shirt sleeve will typically generate enough electric charge to
deflect a stream of water from a tap or pick up small pieces of paper electrostatically. Taking out
a biro to note where a flammable gas is leaking out of a pipe could thus be disastrous.
Fluoroelastomers and PTFE are regarded as satisfactory materials for
oxygen systems because of their high oxygen index. Copper, nickel and their
alloys are regarded as the most satisfactory materials for arduous duty in
oxygen systems, such as pumps, valves and other moving parts, and where a
finely divided material is needed, such as in filters.
Another means of measuring the sensitivity of materials to ignition in
oxygen systems is the autoignition temperature. Finely divided samples are
slowly (200C per minute) heated in a heavy-walled reaction vessel at a
standard pressure, often 100 bar, until they ignite spontaneously. Most
polymers ignite under these conditions at between about 150 and 350C,
although PTFE does not. This sort of test gives a similar ranking to the
oxygen index in most cases.
2.5.8 Safe working in flammable atmospheres
It is possible, and often essential, to be able to work in areas where there
may be a flammable atmosphere continuously or intermittently present. If it
is necessary to work in an atmosphere containing a flammable mixture of
hydrogen, precautions against accidental ignition must be stringent. The
statutory requirements for flammable atmosphere working (i.e. Zone I
requirements) were therefore drawn up with this need in mind.
In a Zone I area all high-power electrical apparatus must be completely
isolated from the gaseous atmosphere. It should be either hermetically
sealed or well-sealed and then pressurised continuously from within by an
inert gas such as nitrogen. Low-power electrical apparatus may be either
isolated or intrinsically safe. The latter is achieved by ensuring that the
maximum voltage in the circuit is only a few volts and that the maximum
instantaneous current possible is in the milliamp region or below. Small
electronic devices, for example, may be used if they contain no large
capacitors and have batteries of a high internal resistance. (Caution must be
sounded over electric torches and flashlights, however; on these the bulb
must be isolated, as the 2500C+ tungsten filament in it would be a splendid
source of ignition.) Barrier circuits are available, using Zener diodes and
resistors, that limit the voltage and current that can be transmitted into the
zoned area along wiring.
GAS TECHNOLOGY 273
All metal surfaces should be electrically bonded together so that ignition
effects from static discharges or electrochemical currents can be avoided.
Plastic materials should be of the slightly electrically conductive anti-static
type. Finally, the use of ferrous tools may need to be controlled. Striking of
ferrous alloys on other metals, stones and other hard materials can lead to
pyrophoric particles. These are microscopic particles of iron which ignite
spontaneously in air or oxygen and can easily lead to ignition of a flammable
mixture.
2.5.9 Gas release accidents
Burning gas leaks. Gas suppliers should always be consulted as to fire
precautions, along with local fire service officials if appropriate. Different
gases will require different approaches, as will different sizes of installation.
However, some general principles are perhaps useful to know.
It is not infrequently the case in large and extensive pipeline systems that
small leaks of flammable gas occur. A small burning leak, especially if it is
outside or in an accessible place inside a building, is often not particularly
dangerous and if prompt action is taken it can be easily extinguished before
real problems occur. First, the gas supply must be shut-off, and, if possible,
the pipeline depressurised via a suitable vent stack. The fire can then be
extinguished in the way appropriate to the circumstances of the leak. It
should be pointed out, however, that extinguishing a gas leak fire, if the gas
supply is still on, is dangerous. If the fire is successfully extinguished, gas will
carry on accumulating and will mix with air, presenting the risk of an
explosion much more dangerous than the original gas leak fire.
Hydrogen installations are particularly subject to leaks, because leaks
very readily, and installations such as thin membrane purifiers and arrays of
many hydrogen cylinders offer many opportunities for leaks. Hydrogen
installations are often placed outside where the impact of a small leak is
normally very small. One of the difficulties of leaks is that they are
exceptionally difficult to see. Outside in bright sunshine the flames are
almost invisible. Old fashioned manuals advise the use of a piece of paper or
a rag on a stick as a flame detector. With glowing carbon particles and atoms
added in this way, a hydrogen flame is indeed very much more visible.
Detonation, deflagration and BLEVEs. The effect of an ignited gas cloud
depends on the speed at which the flame front travels. Gas/air mixtures can
burn rapidly emitting copious heat but causing little immediate mechanical
damage. BLEVEs (burning liquid expanding vapour explosions) are a
special case of this and are typically seen where a container of flammable
liquefied gas such as propane is ruptured and ignited. The mechanical
damage seen is relatively slight, extending only a little beyond the zone of
the visible fireball. In these cases the flame front in the mixture of gases
travels slower than the speed of sound so heat released by the ignition is
dissipated smoothly in the expansion of the mixture. This situation is
referred to as a deflagration explosion. As the burning mixture approaches a
stoichiometric composition, the speed of the flame front will increase and a
deflagration explosion becomes more dangerous. If a sealed vessel ful l of
such a mixture is ignited then the vessel may well be ruptured by the
adiabatic expansion.
41
Much rarer are those accidents where a large volume of near stoichiome-
tric mixture is accumulated before ignition occurs. In these cases, the flame
front travels faster than the speed of sound and, with or without a
containment vessel, there will be detonation explosion. Large volumes of
gas can be burnt and high temperatures and pressures accumulated before
this energy can be dissipated by expansion of the surrounding air. As a
result, the energy of the burning gas is efficiently converted into an
expanding shock wave of immense destructive power. Military high
explosives are designed to produce this kind of detonation explosion.
Military devices (fuel-air bombs) have been devised to utilise the power of
gas-air explosions. Gas burner and boiler manufacturers, and fuel gas
suppliers are often equipped with special explosion test chambers. These are
used to investigate the effects of small detonation and deflagration
explosions as part of their development efforts towards safer burner and
boiler design and use.
Oxidant accidents. Two of the common oxidants in industry are relatively
harmless gases: oxygen and nitrous oxide. Unlike reactive oxidants, such as
hydrogen peroxide or fuming nitric acid, both the gases are stable when
mixed with many oxidisable materials. This makes them in some ways more
hazardous than the obviously perilous nitric acid because they can leak out
in large quantities before ignition occurs.
Liquid oxygen (LO) has more potential than other oxidants for big
accidents because it is used in such large quantities on a routine basis. LO is
easily soaked up by organic material, forming a low explosive.
42
A typical
LO/organic mixture will release more energy per unit mass than an
explosive, e.g. combustion of
nitroglycerine =
octane + mixture =
41
If the vessel is vented a great rushing sound, like a jet engine, will be heard as a very rapid
expansion takes place, followed by the expulsion of a sudden spout of hot gases.
42
For example, one of the party tricks performed by schoolmasters used to be the soaking of a
cigarette in liquid oxygen. An ordinary Benson and Hedges King Size, normally smouldering
for 10 minutes or so, is gone in two seconds, showering sparks everywhere like a firework.
Given t hi s behaviour from a few grams of material, it is clear that the contents of a 50 tonne
liquid oxygen t ank soaked in an organic substance is a major hazard.
GAS TECHNOLOGY 275
The flammability limits of inflammable substances with oxygen are wide,
ignition energy and temperature is low, flame temperature is high, and flame
speeds are high. These effects all conspire to make a fire or explosion in the
presence of oxygen both likely and serious.
Lubricating grease or oil should not be applied to compressed oxygen gas
systems. The density of oxygen at 230 bar is only about a quarter of the
density of liquid oxygen and high-pressure gaseous oxygen tends to dissolve
and form an intimate mixture with organic materials. A typical compressed
oxygen failure occurs when an oxygen system becomes contaminated with
an organic compound such as lubricating grease. The grease may have been
applied with good intentions, for example, to a sticking valve. However,
high-pressure oxygen dissolved in the organic consitutes an explosive,
possibly even a detonating explosive. A small but very dangerous explosion
then occurs, which will rupture the pipework system and perhaps fire out
small parts, such as valve components, like bullets.
It is essential not only not to use lubricating greases and oils in oxygen
systems but also to eliminate any adventitious organic compounds by a
rigorous cleaning procedure for all components used in oxygen service.
There are standard degreasing/cleaning procedures for oxygen components
defined in the national and international codes, and these should be
followed. It is worth noting that cleaning of a complete system, after
assembly, may well be impossible. There may well be dead-legs, parts of
cylinders and loops of pipe which cannot be degreased in situ. In the event
that a system is contaminated with oil or grease, it may well be necessary to
dismantle and degrease each component, followed by reassembly.
As noted in section 2.4.2, the composition of seals in oxygen systems is
also problematical, with perfluorinated polymers most suitable, at least for
less extreme service. Asbestos and related composition seals are often
specified for oxygen service as these are essentially non-combustible, at least
if they are not contaminated.
A much rarer failure occurs when the gas velocity of oxygen down a pipe is
very high, perhaps because of a serious leak somewhere else. In this case,
iron or steel pipes can be ignited, and once ignited will burn and melt in the
stream of oxygen. The pipe then behaves exactly like a thermic lance and can
cause great damage. To avoid this possibility, oxygen pipes should be made
from brazed copper or high-alloy stainless steel, such as 316L, or restricted
in pressure and flow rate. Even stainless steel is not recommended above
70 bar. An accepted rule of thumb is that for steel pipes at lower pressures,
below 40 barg, flow velocity should be kept below whilst at higher
pressures flow velocity should be kept even lower, less than All
oxygen valves and fittings for service at higher pressures should ideally be
made from brass, copper or 316L. However, note that local construction
codes should in all cases be obtained and followed in the design,
construction and operation of oxygen systems.
Sources of ignition in oxidant accidents. It is usually not possible to exclude
all ignitable material from an oxygen system. Polymer seals and flexible
hoses, for example, will always have some degree of flammability in
high-pressure oxygen; even Teflon will ignite at high enough pressures with
sufficient ignition effort.
It is surprising, when a survey of investigated oxidant accidents is made,
how many ways there are of generating sufficient heat to give ignition. The
most unlikely mechanisms have been implicated; for instance, heat gener-
ated by particles of rust or dust in the pipe hitting an obstruction has been
shown to be sufficient under adverse circumstances in a high-pressure
oxygen pipe. The frictional heat generated by turning a valve in a tight stem
seal has also been shown to yield enough heat; these sorts of accidents are
particularly unpleasant because the operator turning the valve may well be
injured. It should be remembered that many powdered materials are
pyrophoric in pure oxygen. Just as tiny particles of misch-metal from a
lighter flint ignite in air as they fall through it, so minute steel particles, for
example generated by abrasion, can ignite in oxygen. Electrical or
electrostatic sources of ignition are commonly blamed for igniting volatile
organic vapours in air and although such sources should have been
eliminated from contact with oxygen systems by design they can occasionally
cause accidents.
Adiabatic compression ignition in oxygen systems. The most common
source of ignition in oxygen systems at high pressure, however, is probably
adiabatic compression heating following rapid opening of a valve onto a
restricted low-pressure system. As discussed in section 2.4.7, this effect
occurs because a small volume of low-pressure gas is sufficiently confined
(perhaps it is open to the atmosphere but only through a relatively small
orifice) that when high-pressure gas is suddenly admitted, the low-pressure
gas is compressed, with a flow pattern that in narrow pipes does not allow
much mixing, unlike the case in cylinder filling. Whereas in cylinder filling
transient temperature rises of a few hundred degrees or less are typical, in
pipes the ful l 1000C or more calculated above might be seen.
In any oxygen system with a high-pressure source, valves opening flow
from that source should ideally be slow-opening screw-jack operated seat
valves; cylinder valves are obviously suitable in this regard. Valves with
particularly slow-opening characteristics (typically by an extended plug,
somewhat like a needle-valve type) are available which will guarantee
particularly slow opening.
The mode of operating of a manual system (or, equivalently, the software
in an automated valve system) can also provide protection against adiabatic
compression ignition problems. For instance, procedures can ensure that all
valves which may have high pressure behind them are opened slowly or parts
of the system which include flexible hoses are always maintained at low
GAS TECHNOLOGY 277
Figure 2.66 Design considerations in a small oxygen installation.
pressure. However, it is most unwise to rely on procedural protection; it is
far better to design a system which cannot easily be put into danger. An
example will show the sort of design considerations that apply (gas suppliers
as well as national guidelines and regulations such as ASTM or CGA should
be consulted before designing or modifying oxygen systems which involve
any high-pressure parts).
In Figure 2.66 a single oxygen cylinder or bank of cylinders is used to
provide relatively high pressure oxygen at two pressures to two user stations.
High flow ball-valves are specified for V1 and V2 so that large flow rates can
be catered for. System 1 will be safe in most circumstances; opening Vc will
not cause adiabatic heating in the diaphragms of the regulator because it is a
slow-opening type. Opening V1 and V2 does not cause a problem because
the pressure (< 30 bar) behind them is unlikely to be high enough to cause a
problem at the user station. In the event of either of regulators Rl or R2
leaking into the downstream (low-pressure) side when V1 and/or V2 are off,
the pressure relief valves PR1 or PR2 will lift, thus assuring low pressure
behind V1 and V2. V1 and V2 could be paralleled by screw-operated plug
valves which have to be opened (slowly by their nature) before opening V1
and V2 for high flow rate, providing protection against adiabatic heating in
the user device.
System 2, although using the same components, has the potential for an
oxygen ignition accident. If Vc has been opened and V2 is then opened
suddenly, high-pressure oxygen from the cylinders and the pipeline to V2
will compress the oxygen just before R2, and R2 may ignite or even explode.
Other rules have evolved from the gas companies long experience with
making and using oxygen systems based on the science explained above, for
instance:
to limit oxygen gas velocities pipes should be generously sized and places
where velocity is artificially increased over the mean, such as at throttling
valves, must be made of inert materials;
high-speed rotating equipment should be designed with positive clearance

under all service conditions to avoid frictional heating;
pressures used at all stages in a system should be minimised.

Nitrous oxide and other oxidants. In general, nitrous oxide is a very safe
gas to handle, with fewer problems than the more ubiquitous oxygen. It does
show all the oxidant accident potential of oxygen, however, with a lesser
degree of reactivity giving a greater safety margin. (However, it should be
noted that some rubber materials are decomposed by nitrous oxide, so
caution must be exercised with seals.)
In addition to its oxidant properties, nitrous oxide has been known to be
the cause of accidents due to thermal decomposition. Occasional failures of
this type happen as follows. A section of pipe carrying nitrous oxide is seen
to be hot, perhaps for no apparent external reason. This section of pipe
quickly becomes red hot and then fails, releasing a large volume of nitrous
oxide at very high flow rate, together with a shower of molten or burning
metal. This sort of incident can occur because nitrous oxide, like acetylene,
is an endothermic gas, i.e. it has a large positive heat of formation (see
Kaye and Laby, 1995). When it is decomposed at a few hundred C into its
constituent nitrogen and oxygen, it releases heat, which causes further
decomposition. The released oxygen may result in combustion of pipe
material, further exacerbating the situation.
Finally, it should not be forgotten that other reactive gases have the
potential to cause oxidant accidents. Fluorine is such a powerful oxidising
agent that most ordinary substances spontaneously ignite or explode. Only
specially passivated stainless steel systems are normally suitable and gas
suppliers specific recommendations for handling this exceptionally hazard-
ous gas should be adhered to precisely. Nitric oxide and nitrogen dioxide are
both relatively unstable compounds, easily yielding their oxygen atoms in
combustion reactions. The other commonly found oxidant gas, chlorine,
also has the power to support vigorous combustion and, furthermore,
combustion in chlorine generally leads to the formation of copious
quantities of hazardous hydrogen chloride gas.
Reactive gas accidents. Purging of all reactive gas pipelines is essential,
with effective venting during purging. Ideally, purging is by a pressure or
vacuum cycle (section 2.5.10), with care taken in venting of the purged gas.
GAS TECHNOLOGY 279
Failing this, purging for a much longer time by passing an inert gas flow will
usually be acceptable from a safety point of view.
43
Silane plants and other dangerously pyrophoric or explosive gas plants are
commonly constructed in friable buildings. Operators observe from behind
solid walls with heavy armoured glass windows. The outside walls and roof
of the building are light, however, to allow any pressure waves to escape
outwards and reduce damage to the operators room. Ultrapure silane, after
its distillation step, is particularly dangerous because it ignites spon-
taneously only occasionally. It can therefore build up to a large volume
before ignition, often explosive, takes place. Frequently, the area around
potential leak sources is exhausted into a duct where there is a continuous
source of ignition, e.g. an electric spark or flame. This burn box then
guarantees that ignition will occur early during the development of a leak, so
that any explosion which may occur will be very small. If a large flame is
used, the burn box also acts as a scrubber, oxidising any suitable reactive gas
present to less harmful oxides.
Helpfully, leaks of many reactive gases can be easily located because of
the reactivity of the gas. Many reactive gases react with moisture in the air to
produce a mist of particles or droplets (the boron trichloride mentioned
earlier is an example). Small leaks of acid gases, such as hydrogen chloride
or sulphur dioxide, can be easily seen if a sponge soaked in ammonia
solution is brought near to the leak; a tell-tale white smoke will appear.
2.5.10 Safe equipment design
Materials selection for gas pipelines and vessels was dealt with in sections
2.4.1 and 2.4.7 but it may be worth saying something in addition about
materials for cryogenic service with respect to safety. Many properties of
materials are better at low temperatures but some are not. The problems of
materials selection for cryogenic use boils down to two prime consider-
ations:
will the material become undesirably brittle at the operational tempera-

ture?
will thermal contraction during cool-down from room temperature match

other components or, if it will not, has the component been designed to
allow for this?
Thermal expansion is more serious in cryogenics because of the simul-
taneously increased stiffness and brittleness of materials; greater strains are
induced and these strains are more likely to lead to failure. Whereas even
43
For example, boron trichloride hydrolyses with moist air to form a dense white cloud of boric
oxide and hydrochloric acid droplets. On purging with nitrogen a very large cloud of
unpleasantly acrid white smoke will appear before the nitrogen purge finally exits the pipe.
Figure 2.67 Elasticity of polystyrene showing glass transition from glass to tough polymer to
rubber and liquid states.
serious mismatches would be unimportant in most applications where
temperatures are cycled 200C upwards, this is not necessarily the case for
200C downward cycles.
Metals with body-centred cubic (BCC) structures are in general unsuit-
able for cryogenic service, as they become as brittle as glass long before LN
temperatures. BCC metals include, unfortunately, most grades of steel.
Carbon and many alloy steels are serviceable in cryogenics only if very
lightly loaded because of these problems and are only used if economic
factors are overwhelming. Consequential failure due to cryogenic embritt-
lement of steel or other parts because of a cryogenic spill should always be
borne in mind in confined layouts. A small LN spill or leak may lead to
excessive cooling of the outside of nearby parts under tensile stress (pipes
under pressure, steel beams or cables, motor or compressor castings, for
example) leading to their failure in service.
Similarly, most plastic polymers become very brittle at a little below room
temperature. The movement of one polymer chain past another, which
accounts for the flexibility and toughness of polymers, becomes increasingly
difficult as temperature is lowered, until the polymer becomes a rigid solid
with a random arrangement of atoms, i.e. a glass. The so-called glass
transition of most useful plastics is below room temperature but only a little
below for most. PTFE is useful to the lowest temperatures (Figure 2.67).
Polymers also have the problem that they have large coefficients of
thermal expansion and poor thermal conductivity. This means that rapid
cooling, for example dunking them in liquid nitrogen, can result in them
undergoing differential thermal expansion leading to distortion and
cracking. Polymer adhesives frequently peel off adherent surfaces after
cycling between room temperature and LN.
Safety in pipework. A set of guidelines for safe pipework is a standard part
of the stock-in-trade of industrial gas firms. These guidelines will cover:
GAS TECHNOLOGY 281
labelling: labels should be attached every few metres of pipeline and given
a colour code if possible;
end-fittings: ideally these should be selected to match unique gas cylinder

or tank fittings so that the wrong gas cannot be connected;
location: location of pipework in the ceiling of a factory, for example,

generally leads to warming of the gas, which may be useful when the gas
has come from a cryogenic source; accessibility for fixing leaks should also
be a consideration for joints in the pipework;
material suitability for gas;
material suitability for pressure;
suitability for contaminants in gas: e.g. particulates in silane leads to the

use of tied diaphragm regulators so that blockage and leakage cannot take
place with particles underneath control valve so easily;
bending, minimum radius, type of bending machine to be used, anneal

state of material needed;
joining: welding, compression joints, flanges, soldering, grades of solder;
flow rates: e.g. flow rates of high-pressure oxygen through steel pipes
allowed are limited.
It should not be thought that all of these guidelines are completely hard
and fast but any deviations from these mostly well-thought out and
time-honoured rules should be carefully justified and agreed with everyone
concerned.
Purging. Purging of a gas system means the replacement of the initial gas
mixture (often air) in a container or pipeline network with the correct
process gas at the highest possible purity. The purging of flammable or toxic
gas systems is necessary for safety reasons. There are three basic ways of
purging:
flow purging: simply passing the process gas through the network;
pressure purging: the system is repeatedly pressurised and then let down
to a lower pressure (typically atmospheric);
vacuum purging: the system is evacuated then backfilled with gas,

sometimes repeatedly.
Flow purging is the least efficient because gases do not flow as perfect
plugs; there is always a slow-moving boundary layer. Also, any dead-legs in
the system will act as long-term reservoirs of impurity. Pressure purging and
vacuum purging deal with these problems, with vacuum purging being by far
the most efficient (Figure 2.68).
Purging theory follows a simple harmonic progression. If the maximum
pressure is P
+
, and the lowest pressure is P
, then the concentration of

impurity C will reduce from an initial in the following way, for n purge
cycles
Thus a vacuum purge of 1/1000 bar is equivalent in efficiency to pressure
purging to an impracticable 1000 bar. With 1/1000 bar vacuum and 10 bar
process gas pressure, the concentration of will decrease by a factor of
10 000 for each cycle in this simple theory. With a simple 1 bar vent each
time instead of vacuum, will go down only by a factor of 10 for each cycle.
The harmonic progression of purging depends on perfect mixing of the
added gas to the lower pressure gas in the pipeline. Clearly this will not be
the case if pressure cycling is carried out from one end of a long pipeline; the
dead-leg end of the pipeline will behave partially as a plug flow and simply
expand and contract as the single-ended purge proceeds. If a line can be
pressurised from one end, and vented from the other, purging will be much
more effective and better than a perfect mixing assumption.
Naturally, practice is more complex than this theory indicates. For
example, if the impurity is water, this tends, at ordinary temperatures, to lie
in substantial amounts in a thin film over the inside of a pipeline or cylinder
and will only slowly evolve under the influence of heat and vacuum. Heat
and vacuum must therefore be used on cylinders for ultrapure semicon-
ductor gases where ultralow moisture content is needed.
Bursting discs and pressure relief valves. No gas system can withstand
infinite pressure and in fact most systems cannot withstand the approxi-
mately 1000 bar pressures which liquid nitrogen can exert if trapped in a
dead-leg and evaporated, so relief valves or bursting discs are essential.
For high-pressure systems, bursting discs (thin diaphragms of a cor-
rosion-resistant metal) can be fitted into special mounting jigs. These act as a
weak link in the system, bursting before any major component can be
damaged and releasing the gas pressure. Only a small amount of debris is
ejected when a disc blows; an important safety feature, although the event is
spectacularly noisy and not a little frightening for nearby spectators.
Figure 2.68 Purging a gas line of air by vacuum and pressure.
GAS TECHNOLOGY 283
Bursting discs are carefully stamped out of metal sheets, with a testing regime
to ensure that they will burst at the correct pressure. Small variations in alloy
composition and in heat treatment or rolling can lead to big variations in burst
strength, so very careful quality control and testing is essential. They are
available only in a limited number of pressure ratings matching common
requirements, as they are economically made only in large numbers. They
should always be mounted in a supplied or recommended holder, as even
small changes in the holder design may affect the effective pressure rating.
Related to bursting discs are the fusible plugs often fitted to cylinders. These
are made of a solder alloy which melts and releases gas at modest pressures in
the event of a gas cylinder being overheated in a fire.
At very high pressures (100 to 10 000 bar), bursting caps or tubes may be
seen. These are more common on hydraulic than gaseous systems because of
the problem of debris ejection. They can be machined out of solid rod material
and made highly predictable. A bursting cap is simply a machined dead-end
with an accurately machined groove around the pipe circumference below the
end, giving a weak point of known width and thickness. The end cap detaches
on bursting and must be retained, however. A machined piece of thick-
walled tubing with an accurately made thinner walled section can also be
employed, which splits axially on overpressure. In addition, at the highest
pressures, bursting discs may be seen which are thinner around the edges than
in the middle. Provided this is done carefully, it offers the advantage that the
disc will reliably burst leaving a completely clear flow path and maximum
venting flow (e.g. Anon, 1975).
Lower pressure systems, and higher pressure systems which need a
resettable pressure relief system, employ spring-loaded check valves. The
standard type of pressure relief valve is related in its mechanical design to the
check valve described in section 2.4.2. In it a spring holds down a seat-type
valve; in the event of excessive pressure (more than the cracking pressure)
the seat lifts and the gas is vented. Simple relief valves reset when the internal
gas pressure falls to a little below the cracking pressure, allowing the spring to
close the seat valve. Other relief valves require a reset pin or button to be
pushed to reseat the valve; this is only possible when the internal gas pressure
has fallen.
Pressure relieving devices of some type should be employed:
where cryogenic liquids might become trapped;
where chemical reaction or increase in temperature might cause a large

increase in pressure in a trapped volume;
where the failure of a pressure regulator might result in overpressurisation

of the system downstream of the regulator;
wherever there is a large vessel or pipe whose failure would release a large
amount of pneumatic energy;
on systems used to contain a high-pressure liquefied gas such as

Relief valves must not only be set to relieve at the correct pressure, they
must be sized to allow a reasonable relief flow rate. In most cases the
manufacturers give flow curves but approximate calculations are simple and
reasonably accurate on a standard orifice plate basis (see, for example,
section 1.3.8) for many designs of bursting disc or valve.
For systems which may rise rapidly in pressure, reactor vessels with rapid
exothermic reactions, for example, the size chosen must be very generous,
as reactions generally increase in rate with rise in pressure and temperature.
A rapid build up of P and T thus leads to a thermal runaway and bursting of
the reactor vessel if relief is not rapid. For other systems, a cryogenic tank is
an example, where the maximum possible required relief flow rates are more
easily calculated and in any case more modest, much smaller relief valves or
bursting discs are appropriate. An overlarge bursting disc could itself lead to
problems, especially as the bursting disc, unlike the relief valve, will not
reset after a while but will allow all the gas in a system to escape. The huge
cloud of freezing fog produced when tons of LN is released suddenly, the
possibility of an oxidant accident with a sudden massive release of oxygen,
and the toxicity of a large release of carbon dioxide are all examples of where
a bursting disc actuation could itself cause a problem.
Sizing of pipes leading to pressure relief valves must always be generous.
If possible, relief devices are located immediately adjacent to the largest
vessel or pipe in a system. Consideration must also be given to where the
relief valve discharges to. Ideally a tall, large-diameter vent stack discharges
the relieved system into the open atmosphere several metres above where
any human might possibly be.
Gas storage location. Gas cylinder location and installation should never
be arbitrary. Toxic and flammable gases can often conveniently be located
outside along with a primary pressure regulator, a rain shelter above and
low-pressure gas piped inside to the point of use. Any leaks during cylinder
changeover are then easily diluted in the open atmosphere. Cylinders should
in any case be secured so that they cannot topple and precautions such as a
guard or cap should be taken to ensure that the cylinder valve cannot be
sheared off in the event of it being dropped.
Gas cylinders on vehicles are another area where safety must be
considered. Gas cylinders should not be carried in the same compartment as
the driver and passengers. A modest leak of gas could lead to incapacity of
the driver and a vehicular accident. Alternatively, a vehicle accident could
result in valve rupture or leak and present an additional hazard to the
occupants. The elevator or lift in a building is equally a vehicle from these
points of view, and cylinders and cryogenic containers should be sent up in
the service lift without a human attendant if possible.
Empty cylinders can still contain a significant amount of gas under
considerable pressure. One reason is increasing moisture level at low
GAS TECHNOLOGY 285
pressure which prevents the use of the last 20% or more in high-purity
applications. Another problem is the minimum pressure required to run a
particular system. In any case, a declared empty cylinder may well turn out
to be ful l ; there is little difference in weight, so unless a pressure gauge is
applied and the cylinder valve opened, it is difficult to be sure that a cylinder
is empty. Empty cylinders should therefore be accorded the respect paid to
ful l cylinders.
2.5.11 Safety audit and HAZOP/HAZAN analysis
Every system in regular use for whatever purpose has a probability of
failure. Sometimes this probability is exceedingly small, or the conse-
quences relatively light, and this risk is simply accepted. More often the
failure probability is high enough, or the consequences sufficiently serious,
that the system is designed to include back-up components. Alternatively,
the system is designed to fail-safe, in other words to fail in such a way as to
avoid serious consequences; reliability is impaired but not safety.
Complex systems can end up being surprisingly unreliable and unsafe if
insufficient thought is given to the probability and effect of failure of
components and combinations of components. For example, suppose that a
liquid oxygen burner system had 50 components that were all individually
critical to its safe operation; components like perhaps the tank pressure
regulator valve, the vaporiser coil, the burner cooling water, the flame
failure sensor, and so on. Further suppose that each of these fails randomly
and has an MTBF (mean time between failures) of 20 years, which might
seem reasonable if a system lifetime of 10 years is needed. The MTBF of
such a system would in fact be only a few weeks. If even a small number of
the failure modes were unsafe, this would be unacceptably bad perform-
ance. (Real systems will often, but not invariably, be better than this
because many failures are wear related and have a more or less definite
lifetime; put another way, their MTBF in their early years is much lower
than their average MTBF.)
The reason for the bad system MTBF in the above case is the result of
multiplying together the probability of no failures. This magnifies even small
deviations from perfection because any one defect anywhere in the system
causes complete failure.
For small component failure rates, the system failure rate is
simply the sum of the component failure rates This kind of
design is, of course, avoided in practice, and systems are designed where
possible to provide, for example, a back-up subsystem. Provided the
back-up subsystem is truly independent of the primary, and this must be
proved beyond reasonable doubt, then the probabilities of failure are
multiplied and the system failure rate reduced
For small failure rates, the MTBFs for the subsystems are added.
Hazard/operability study: HAZOP. The basis of HAZOP is a systematic
account of all the possible failures of individual components or subsystems.
This essentially qualitative method can then be followed by quantitative
calculation, calculating the probability for consequent failure modes from
estimates of individual component failures; this is then HAZAN (hazard
analysis).
HAZOP has proved to be a valuable discipline for bringing together and
formalising the thoughts of the designing engineers and scientists, and the
operators of gas systems with respect to safety and efficient operation.
Particularly when a new design of plant is being designed, the right format
must be arranged.
Correct group ofpeople. It is important to collect together the right
group for HAZOP. It must include the designers, the users (including the
shop-floor users), the scientists who developed the new process, and an
independent chairman, perhaps a delegate from the safety department or a
consultant. The best teams probably include creative thinkers with little
detailed knowledge of the plant, as well as those closer to the project. In the
case of applications of industrial gases, those who work for the gas user
should be involved, as well as gas company personnel.
Correct level of authority. It is a waste of time if the HAZOP group does
not have the necessary authority to implement at least small changes in plant
and its operation.
Correct orientation and guide words. The starting point for HAZOP is to
think up all the possible ways the plant can go wrong. To do this, it is helpful
to use a set of guide words such as are listed below. These words are then
applied to the parameters of the gas system in question. The question What
could go wrong if . . . ? should be asked for each parameter:
None e.g. pressure/flow/reverse flow/lubricating oil
Part of e.g. omission of part of gas mix
Less of e.g. lower temperature, flow, pressure
More of e.g. lower temperature, flow, pressure
More than e.g. impurity or air/water/oil inclusion
Other than e.g. other than normal mode, such as shutdown,
startup
GAS TECHNOLOGY 287
In each case, the interaction of any abnormalities with operator action must
be considered. Some abnormalities will be obvious to the operators, others
will be measured and relayed to them by specifically provided instruments,
whilst others may occur incognito.
The application of these principles must then be applied at selected points
throughout the plant under study. It is of course a matter of judgement how
far apart those points lie. However, points separated by reliable intervening
components, valves, regulators, etc., are probably a reasonable guideline.
Hazard analysis: HAZAN. There are essentially three elements in any
quantitative study of hazards: failure probability, damage consequent on
failure and a judgement on the acceptability of the average consequences,
i.e. the product (failure rate) (consequent damage). Each of the failure
mode probabilities in a HAZOP may have been measured. However,
failure data is subject to statistical error, with only relatively small numbers
of failures typically having been observed. Also, rare failures may not have
been observed at all but must be inferred and probabilities estimated. For
these reasons, failure rates are always subject to considerable error.
Even bigger errors creep into HAZAN, however, when the consequences
of failure must be considered. How big will the leak be when it happens?
How far will the fire spread? It is almost impossible to answer these sorts of
questions quantitatively from experience but estimates must be made to
complete the analysis.
Take the example of a pressure vessel safety relief valve. Suppose a relief
valve is tested every year. Further, suppose that on average, 1 in 100 relief
valves tested has failed stuck down, i.e. failed dangerous, as opposed to
failed safe, and that the pressure vessel it relieves goes overpressure every
three years. The rate of hazard occurring on the pressure vessel as a result of
this is therefore 0.3 0.5 x 1% every year, i.e. 0.0015 hazards per annum or
about one accident every 700 years. (The valve is likely to fail half-way
between tests on average, i.e. it will be failed for 0.5 years before discovery
on average. The vessel has a 1% chance of having a failed valve on it each
year and it tries the valve out 0.3 times per year. A more complex formula is
appropriate where the test interval exceeds the interval between tries.)
It is not sufficient to stop at this point and give the relief valve a clean bill of
health, however. Further calculation is indicated. Suppose the vessel
concerned contains a harmless compressed gas but the rupture would almost
certainly kill or badly injure the operator working with a machine
incorporating the pressure vessel. The fatal accident rate (FAR) resulting
could be as high as 15, i.e. probably higher than would be acceptable in the
gas industry, which is aiming at FARs in planning of 0.5 or less. The relief
valve would therefore need to be paralleled with another similarly reliable
device, preferably of a different technology, for example a bursting disc.
Finally, at the end of a HAZAN, the acceptability or otherwise of the
average consequences must be decided. This area is curiously subjective. It
has been estimated, for example, that the nuclear industry spends 10m or
more for each life saved on safety equipment, whilst the healthcare industry
finds it difficult to justify spending more than 100k or so to save a life. The
FAR is defined as the number of fatalities resulting from each working
hours. The FAR of many industrial activities is below the FAR involved in
ordinary life (in which many fatal accidents are due to motor vehicle use),
which is effectively around 20. Only a few workers, such as fishermen,
offshore oil workers and building erectors, have a FAR significantly above
20.
In the past there were many more dangerous industries. Today dangerous
activities are not tolerated, at least while it is reasonably easy to make them
safer, which appears to be possible for all but exceptional work such as
offshore oil production. Fortunately, the gases industry in most countries is
reasonably safe, with FAR in the 110 region. It is also worth remembering
that many accidents in the gases industry are in fact road accidents while, as
remarked above, the most common accidents in the gases workplace are
unremarkable incidents during handling such as dropping a gas cylinder on
the toe or cuts on the hands. Some simple safety equipment such as gloves,
goggles, safety spectacles and safety shoes with steel toecaps are very
worthwhile to minimise risks.
2.5.12 Environmental issues
The stratosphere is defined as beginning at 10 000 m. While the atmosphere
below, the troposphere, contains most meteorological phenomena and
contains most of the mass of the atmosphere, the stratosphere is important,
too. Although the stratospheric air is very thin (only 1/4 bar at 10 000 m and
l/100th bar at 30 000 m) the sunlight which reaches us is filtered by
substances in it such as ozone. It is only comparatively recently, however,
that the importance of the stratosphere has been realised.
Ozone in the stratosphere. Ozone exists in the stratospheric layers and
absorbs most of the ultraviolet (UV) radiation produced by the sun. Without
this reduction in UV light, widespread skin cancer would occur. There had
been worries for some time amongst atmospheric scientists that the
stratospheric ozone might be affected by man-made gases released into the
atmosphere, especially chlorofluorocarbons (CFCs). The publication in
1985 of results of the British Antarctic Survey confirmed the worst fears,
showing that for five successive springs Antarctic ozone had been plummet-
ing to unprecedented low levels at two locations; the so-called ozone hole
(Figure 2.69).
There is now comprehensive evidence that ozone holes are forming above
the polar regions. Some of the Nimbus (see Houghton et al., 1984) series
GAS TECHNOLOGY 289
satellites have mapped ozone distribution thoroughly and more monitoring
stations on the ground are filling in details in the pattern of change. The
ozone holes have been predicted to spread to lower latitudes, according to
models of ozone depletion which have been proposed, and there is already
evidence that this is happening.
Ozone depletion is primarily caused by halogen atoms causing chain
termination of the ozone formation reactions. Photodissociation of a CFC is
followed by reaction of the chlorine atom with ozone. Reactions such as
occur. The latter reaction recycles the Cl atom to cause more ozone
destruction and destroys an oxygen atom that might otherwise go on to form
ozone
The CFCs lead any list of ozone-depleting substances. Their estimated
relative effects (ozone depletion potential, ODP) on stratospheric ozone are
given in Table 2.18. Newer compounds, regarded as replacements for
traditional CFCs, are considerably better than the bench mark CFC-11 and
CFC-12 (Table 2.19).
probably has a much lower impact because it is more reactive than
the other substances listed in Tables 2.18 and 2.19. In sunlight it reacts with
oxygen in the air to form the highly reactive (and toxic) phosgene. It is the
chemical stability of CFC compounds that paradoxically causes them to be
the potent source of ozone destruction. Their low solubility and low
reactivity with water or oxygen in the troposphere means that they are not
washed out there, unlike more active halogen compounds such as HC1 or
HF. Their lifetimes in the troposphere are thought to vary between 65 and
400 years. These very properties, resistance to hydrolysis and low flamma-
bility, are those for which they are valued.
Many industrial spokesmen were vehement in their opposition to the
Figure 2.69 Development of ozone hole over Antarctica (Bridgeman, 1990).
banning of CFCs and vehement that these molecules, with their remarkable
properties of zero flammability and toxicity, and wide range of boiling
points, were irreplaceable. The availability of the new HCFC range of
molecules, also non-flammable and of at least as low toxicity, but with zero
or almost zero ODP and of acceptably low reactivity, shows how groundless
this opposition was. Most applications formerly using CFC compounds can
now use HFCs. Azeotropic or near azeotropic mixtures of the HFCs are now
available, which has further extended the range of available boiling points.
and VOCs. (nitrogen dioxide and its dimer dinitrogen tetroxide,
nitrogen monoxide and nitrous oxide) is formed during high-temperature
combustion processes. VOCs (volatile organic compounds) are anything
from natural tree resins seen near conifers in the summer to petrol vapour
evaporating from vehicle fuel or isobutane from aerosol spray cans.
Ozone is as undesirable in the troposphere as it is desirable in the
stratosphere. It is toxic and causes chronic respiratory ailments even in very
low concentrations. Ozone occurs near ground level in the troposphere. It is
often found in the ugly brown-stained air above industrial areas which are
situated in a natural bowl, especially in clear, sunny, windless weather.
Normally the atmosphere at ground level contains little ozone. However,
the presence of large quantities of and VOCs, in the several parts per
million region, changes this. and VOCs react with oxygen in bright
sunshine to produce ozone at low concentrations. High concentrations of
GAS TECHNOLOGY 291
are themselves toxic, so effort has always been directed against them.
However, VOCs have only recently been recognised as the cause of ozone
formation at low levels. Investigators looking at serious smog problems in the
Los Angeles area first proposed VOCs as the main cause and subsequent
studies now confirm this. VOCs are now rated on the POCP (photochemical
ozone creation potential) scale, with the simple saturated hydrocarbons rated
intermediate between oxygenated compounds, such as alcohols, which have
low POCP and complex aromatics, which have high POCP.
Other atmospheric pollutants. Carbon monoxide is chief amongst other
pollutants. CO is produced during incomplete combustion of organic
material. However, external combustion devices, burners, do not produce
much CO unless they are badly adjusted. Most CO comes from internal
combustion engines. Particularly when it is idling, the car petrol engine
produces CO copiously; depending on the adjustment of the carburation
system, from 1% to 8% CO is common. Although aimed at the catalytic
exhaust treatment units now common on vehicles also reduce CO. Despite
this recent improvement, many countries are now concerned about CO
production from vehicles and are insisting that mi ni mum standards are met.
These new government regulations mean that cars must be tested for CO
and then adjusted to minimise it. This has led to a worldwide demand for
calibration gases for the analysers, often the non-dispersive infrared type,
employed in the CO testing. Special gas divisions of the industrial gas
companies supply CO mixtures for simpler analysers and mixtures with
accurately known contents of moisture, and hydrocarbon vapours for the
more sophisticated multigas (hydrocarbon + CO) analysers. In the UK
currently, for example, the Ministry of Transport vehicle testing stations are
supplied with a mixture containing 2400 ppm propane, 6% CO and 15%
in nitrogen for the calibration of gas analysers.
The greenhouse effect. A greenhouse keeps plants warm on a sunny day by
allowing visible light through its transparent windows to warm up its contents
but does not allow infrared radiation and convection to cool the plants. The
earth can be regarded as a kind of giant greenhouse. The atmosphere of the
earth, although its principal constituents are transparent to visible sunlight,
contains infrared absorbing gas molecules such as and especially
which block the radiation of heat from the surface to space. Emissions of many
man-made gases, especially into the atmosphere are increasing the
greenhouse effect and it is thought to be possible that these emissions are
therefore causing significant global warming.
The emission of carbon dioxide is an intrinsic part of fossil fuel (oil, gas,
coal) burning. The emission of is so copious that no practicable
scrubbing system can be devised. (Suggestions that power station exhaust
stacks could be cleaned up and the resultant nitrogen stream bubbled
through a few square miles of made-to-measure mangrove swamp have to be
regarded as still at the concept stage.)
The effect of increased efficiency in any energy-intensive processes is to
require less to be emitted. Efficiency increases are therefore one of the
few currently effective weapons against excessive emissions. Com-
bustion using for high-temperature industrial processes and furnaces also
has, in general, a positive beneficial effect on emissions, inter alia. This
arises in the following way. Suppose a process needs, say, a temperature of
1500C to operate. This can be provided by fuel/air at 2000C, but in this case
75% of the heat of combustion is wasted in the exhaust stack gases, which
exit at 1500C unless complex regenerators/recuperators or other energy
recovery systems are in use. However, an oxygen-enriched air stream at
2500C would find only 60% of its heat of combustion wasted.
Although most global warming fears are centred around many other
gases are much more potent infrared (IR) absorbers, and could give rise to
problems at very low levels; much lower than the few 100 ppm of that
are necessary to influence global warming. Long-lived IR absorbers that
could in principle be a problem include the CFCs, VOCs and nitrous oxide.
A glance at the list in Table 2.20 of the global warming potential of a
selection of gases, together with the approximate concentrations of the
gases, reinforces the view that is only the principal problem because of
the large volume of emissions.
The correspondence between carbon dioxide and the halocarbons
depends on assumptions about the lifetime of gases in the atmosphere as the
global warming potential is an integral of the warming effect produced over
the lifetime of the gas. It is assumed here that and the halocarbons are
GAS TECHNOLOGY 293
injected at time zero and their effect measured over 20 years. Short lifetime
gases have a disproportionately large effect on these figures. If a longer time
such as 100 years is taken, then the long lifetime gases will have a larger
GWP relative to which is reckoned to have roughly a 120 year lifetime.
The longer integration time is not appropriate because technology and usage
of gases changes significantly over 20 years and will most probably have
changed dramatically over 100 years. However, some of these lifetimes are
so long that current technology may leave an undesirable legacy, as has
happened for CFC-11 and CFC-12 in ozone depletion.
Local pollution issues. Although unimportant on a wider view, there are
potent local pollution problems which still need resolution in many
industries. River and lake pollution is an example of this. Oxygen (section
3.6) and other gases used to control waterborne pollution problems can have
a significant positive impact on local pollution problems.
2.5.13 Industrial gases and the environment
The gases industry itself has been, and remains, in general one of the most
benign to the environment. With few exceptions, industrial gases have
themselves been made in clean processes and have enabled cleaner
processes to be used in the rest of industry.
Examples of where damage to the environment has been reduced by the
use of industrial gases include the following:
solder inerting with nitrogen where CFC solvent use is no longer required
for cleaning off flux residues (section 3.5.14);
oxygen processes to make steel have reduced the formidable capacity of

steelworks to pollute the environment with smoke and dust and reduced
the use of coke ovens.
Less impressive examples, where the gas industry has led to increased
pollution, include the production of acetylene. Calcium carbide production
via electric arc furnaces produces copious smoke, whilst the reaction of the
carbide to make acetylene results in large quantities of lime sludge.
Environmental audit and life-cycle analysis. Environmental audit aims to
determine the ultimate benefit, so-called life-cycle analysis, to the environ-
ment of alternative products and processes. The environmental issues can be
summarised in an analogous way to a financial statement comparing the
monetary pros and cons of projects. For example, the analysis of an aerosol
can versus a pump-spray unit is shown in Table 2.21. From this rough
example, it is not clear, contrary to claims often made, that a pump spray
package has any real overall environmental benefit over an aerosol package.
The results of a careful environmental audit are often surprising in this way.
Ecopoints. With life-cycle analysis it is clearly now possible to compare the
overall environmental impact of options. There then arises the question of
whether any measure can be devised which enables the comparison to be
made quantitative. Is it better, for example, to emit from a furnace and
save the that would have been used if electricity had been consumed to
generate oxygen for running that furnace with low emission? The
Ecopoint scheme, being developed by Dutch and German groups (e.g. Pr
Consultants, Amersfoort, Netherlands) is one of a number of recent
attempts to do just that. It weights energy used and wastes emitted in order
to drive decision-making in a consistent way. Initially, the assignments are as
shown in Table 2.22.
2.5.14 Scrubbing
Gas scrubbers are vital in modern industry. Although more efficient and
intrinsically cleaner processes are making great strides, and gross polluters
like the LeBlanc process have disappeared, many plants have unavoidable
effluent gas streams. The term scrubber in the gases industry has a slightly
wider usage than its meaning of a gas/liquid contacting tower in chemical
engineering and encompasses basically any device, whether using liquid or
solid, which removes a pollutant.
GAS TECHNOLOGY 295
Filters are, of course, employed in many effluent streams; their functioning
in the role of clean gas supply was noted in section 2.4.3. In the effluent role,
different filters come into play to address the different problems posed by
dirty gas. Although efficiency is still important, low pressure drop and ease of
removal of filtrate become equally important in dealing with exhaust gases.
Bag filters can be used to solve the latter problem, together with arrangements
to shut off flow and vibrate the bags to release the precipitate from time to
time. Instead of low-capacity membrane filters, depth filters, with a thick
bed of relatively larger fibres, are more attractive. Electrostatic precipitators
or cyclones are likewise more useful on dirty effluent streams than
low-capacity filters.
However, filters and electrostatic precipitators are sometimes a messy
solution to the problem of particulates in effluent gas streams and no solution
at all for the problem of unacceptable gaseous contaminants. Scrubbers
proper are then required. Some scrubbers actually achieve a completely
benign catalytic conversion. The precious metal catalyst in the motor car is an
example of this; once up to operating temperature, it converts into
nitrogen by catalysing the reduction reaction. In industry there are examples,
too, such as the SCR (selective catalytic reduction) process for the reduction
of in furnace exhaust stacks and the Claus process for removing hydrogen
sulphide.
Wet scrubbers. Scrubbers were used to absorb evolved HCl from the
LeBlanc process in the nineteenth century. Since then, designs have
advanced in execution, though not in basic principle. Wet scrubbers remove
gas impurities by absorbent liquid sprayed down a tower or trickled over
packing material in a column through which the gas stream is passed in an
upward direction. With many impurities, water washing can be used. Many
particulates, acid, alkaline and some organics, can be effectively absorbed
and removed. With other impurities, special solutions, such as ethanolamine
for carbon dioxide or hydrogen sulphide contamination, or sodium hydroxide
for arsine and phosphine, can be employed.
The largest scrubbers in use are those for removing sulphur dioxide from
the exhaust of electric power stations burning coal. These scrubbers typically
use a slurry or powdered limestone reactant, which is converted by the
exhaust into a rather unattractive calcium sulphate/sulphite by-product
(although the latest processes are capable of producing a better sulphate, a
usable grade of gypsum for plasterboard manufacture).
With liquid scrubbers there remains, of course, a liquid effluent disposal
problem. In many cases, the highly dilute solution produced is completely
acceptable in water courses and no further treatment need be applied; this is
open cycle scrubbing. In others, the contaminant gas may be desorbed and
usefully employed (as in stripping from ethanolamine, e.g. Kohl
and Riesenfeld, 1960) and the liquid recycled; this is closed cycle scrubbing.
Wet scrubbers are often combined with another technique. For example,
a wet scrubbing system can be combined with an incinerator technique
whereby the incineration is carried out at a submerged burner. Dry
scrubbers, e.g. incinerators and plasma-based scrubbers are mostly used in
the semiconductor industry.
Dry scrubbers. Dry scrubbers are simple adsorbers for gaseous contami-
nants, typically based on porous carbon. They are popular in the semicon-
ductor industry where small percentages of occasional contaminant are
removed from large ventilation air or nitrogen purge flows, typically with a
specially activated carbon bed. Similarly, in the coating industry, nitrogen
contaminated with solvent can be processed through a carbon bed adsorber
where large amounts of organic solvent can be adsorbed, particularly at low
temperature (Figure 3.36). The bed can be regenerated by heating in a pure
Figure 2.70 Absorber bed system with exchangeable absorber cartridges (courtesy of Edwards
High Vacuum Ltd, Crawley, UK).
GAS TECHNOLOGY 297
gas stream for a while and the solvent recovered. In some systems the
nitrogen or air scrubbed is also recycled.
Other scrubbing/recovery systems are based on using liquid nitrogen or
other means to cool the exhaust gas, causing condensation or enhanced
adsorption of the solvent vapour or gas. The nitrogen added is used as the
carrier gas and recycled. Ingenious use of cold recovery from the gas
stream minimises the coolant usage (see section 3.7).
The capacity of carbon absorbers for low concentrations of impurity is
reduced, as examination of a simple equilibrium equation shows. For low
levels of impurities such as the ppb levels of mercury in amalgam brine
electrolysis cells, it is useful to add reactive chemicals to the porous carbon.
In the case of mercury, for example, the addition of iodine to the absorber
bed blocks mercury for much longer than a simple carbon bed.
Heated scrubbers are also available and these also rely on chemical
reaction with the effluent gas. These can be very simple; heated calcium
oxide (lime), for example, reacts with acid gases to form stable harmless
salts. The low cost of the lime means that it does not need to be recycled.
One system sold for removing semiconductor vacuum pump exhaust gas
claims to remove more than 20 common (mainly acid) gases (see Figure
2.70).
3 Applications of gases in industry
3.1 Gases in metal smelting and extraction
3.1.1 Oxygen in ferrous metals
The biggest air oxygen plants in the world supply oxygen for steel making. A
single liquid air distillation column can produce up to 2000 tonnes or so of
oxygen per day for a steelmaking site. This thirst for oxygen of the steel
making process is prodigious: it is equivalent to the oxygen breathed by over
half a million people. The fact that so much oxygen is used may seem
puzzling at first glance because the basic process of producing steel is the
chemical reduction or deoxidation of iron ore, essentially by adding a fuel
such as carbon
Some oxygen is used for cutting steel (steelworks are the biggest users of
oxygen cutting equipment). However, applications for oxygen such as these
use in total only small amounts of oxygen. Much more oxygen is used in the
oxidation of impurities in iron and much more than that again in blast
furnaces.
3.1.2 Blast furnaces and the smelting of iron
A typical blast furnace is basically a very large continuously fed smelting
furnace (Figure 3.1). Solid raw materials, iron ore, coke, limestone and
various minor additions, are all carefully prepared by sophisticated grinding
and pelletising processes. These solids are added to the top of the furnace via
a complex arrangements of large bell-valves. Also at the top is the gas
exhaust, now not allowed straight into the atmosphere as it once was.
44
The
exhaust is now diverted via various filters and scrubbers to clean up both
particulates and gases, some of the latter being recycled. In some processes
the top-gas from the blast furnace is an important by-product itself. At the
high temperatures (1500C or so) in the furnace, the iron ore is reduced to
44
Only a few years ago this exhaust provided steel towns such as those in southern Wales in the
UK, with spectacular smoke plumes that could be seen from 50 miles away on a clear day and
cast a continuous, stratified, often reddish-hued dark cloud over their surroundings.
APPLICATIONS OF GASES IN INDUSTRY 299
Figure 3.1 Conventional blast furnace ironmaking.
molten iron, which then trickles down through the porous coke to the
bottom. The limestone and other materials react with phosphorus, sulphur
and so forth in the ore and also melt and trickle down. The coke, having a
much higher melting point (3700C) provides the solid substrate for these
reactions.
Down near the bottom, molten iron, alloyed with up to several per cent
carbon from the coke, is tapped off, either directly into steel making or, less
frequently in modern works, into temporary storage by being poured into
pig-iron ingot moulds and allowed to cool. A little above the pig-iron
tapping at the base are a set of water-cooled inlet pipes, the tuyeres, for
blasting the mixture with oxygen, hot air or other blast gas mixtures. Other
tappings allow for sampling of the charge during operation and removal of
molten slag (the limestone and its products of reaction and impurities in the
iron ore).
Standard blast furnace design and operation has to follow a number of
rules, at least approximately, as follows:
The furnace must vary in cross-section, first increasing in diameter from

the fi l l i ng bells gently and then waisting towards the bottom more sharply.
If this kind of variation in cross-section is not employed, then there will be
problems of the charge collapsing on itself and blocking gas flow or,
conversely, the charge will not descend quickly enough during reaction,
reducing the furnace throughput.
The gas flow must be as high as possible to maintain throughput, but not so
high as to blow too much of the smaller pieces of charge out of the exhaust
or so high that the burners heating it cannot maintain temperature.
The temperatures chosen have to be high enough that the oxygen in the
blast can oxidise the coke to CO. The CO then performs most of the actual
reduction of the iron oxide ore to metallic iron. Temperatures must also
be hot enough everywhere that the high-carbon (5% or more) alloy
produced is melted, at least in the lower part of the furnace.
The injection of oxygen provides a valuable additional variable in the
process. Oxygen enrichment of the blast air heating the furnaces enables
higher temperatures at relatively low capital cost, and some increases in
capacity, and has been a retrofit feature in many steelworks upgrades.
Today, however, oxygen enrichment is fitted into new blast furnaces from
day one, a design change which can be regarded as mostly due to the
increasing cost of coke.
Coke is made by the pyrolysis of coal. The special grades of coking coal
needed to prepare coke have become increasingly expensive and have had
their own effect on coke cost. However, in addition to coke, the process also
produces prodigious amounts of noxious sulphurous gases, and new coke
ovens can only be installed with extensive exhaust gas treatment to limit
environmental impact. The result of this is that coke ovens can now cost
more than 30% of the total cost of a steelworks, more than the blastfurnaces
they serve.
There is therefore an increasing emphasis today on reducing the need for
coke in blast furnace operation. One way of reducing the amount of coke
needed is the injection of fuel through the tuyeres, providing an alternative
source of heat and CO gas. Both finely powdered coals or a spray of
hydrocarbon oil can be used, with the former being very attractive for its low
cost and high carbon content. However, this cannot be carried very far
before the heat absorbed by the endothermic pyrolysis of the fuel means that
the flame temperature achieved by the tuyeres is reduced, which is where the
requirement for oxygen comes in.
The use of both fuel and oxygen additions to the blast air is crucial in
allowing large amounts of fuel to be added. The oxygen enrichment raises
the tuyere flame temperature. The overall effect is to allow a much lower
coke/iron ore ratio, reducing coke costs more than the added fuel/oxygen
costs. Fuels may be heavy petroleum oils, natural gas, powdered coal or
coal/oil slurry. Enrichment factors of 5% or so are typical and of
O
2
and 100 kg of fuel per tonne of iron is needed to save 150 kg of coke. With
a 10 000 tonne per day blast furnace, therefore, an oxygen demand of
or 500 tonnes per day is indicated. In fact, oxygen plants of much
greater capacity, up to about 2000 tonnes and often several plants, are used.
This reflects the fact that other uses have been found for oxygen in
steelworks. In addition, the year-on-year increase in the size and number of
Figure 3.2 Bottom of blast furnace showing the ring of tuyeres which inject fuel and
oxygen-enriched air into the burden.
blast furnaces using oxygen and their increasing oxygen usage has led to
plant sites with a more complex supply system than would have been the case
had an entire new gas plant been designed to satisfy current usage.
3.1.3 Increasing prospects for oxygen in blast furnaces
There is ongoing work at conventional blast furnaces to increase further the
amount of injected fuel, decreasing further the need for coke. Coal injection
rates of 250 kg per tonne of iron now seem achievable (see Corbett and
Smith, 1993) and this will need oxygen enrichment of 10% with blast
temperatures of 1100C or so. This kind of continuous evolution will
undoubtedly increase the demand for oxygen in steelworks in the future.
More revolutionary developments, such as ore injection or the carbide
route, may also affect oxygen usage in the blast furnace. The standard blast
furnace cannot directly utilise iron ore. The fines in the ore would block the
furnace when it was loaded at the top or be blown out by the gas flow.
Instead, the ore is ground finely and then sintered, with additives, into lumps
(sinter) which are then broken down into roughly equal size. However, in a
blast furnace steelworks much of the energy usage, and the capital
investment (up to 30% or so), will be needed for the ore grinding, sintering
and pelletising plant. It makes sense, therefore, to t hi nk of adding fines
directly to the furnace, thus saving the sintering and pelletising processes.
Fines can be blown into the tuyeres, where they will be reduced and melted
before they cause blockage or loss problems.
Beyond the use of blast furnaces, it is quite possible that the new processes
Figure 3.3 Simple schematic of the COREX process.
under development, which are moving away from the blast furnace, may
become significant in the future. Some of these possibilities are:
DIOS: this is a Japanese project, funded by the central government

agency MITI, to use ore fines, coal and pure oxygen all directly; a 500
tonne per day pilot plant is now being completed by the Japanese
company NKK at Keihin.
CCF: the Cyclone Converter Furnace is being developed by a British

Steel/Hoogovens/Ilva/CSM consortium, using a two-reactor approach; an
upper cyclone reactor using ore fines for preheat, reduction and
production of an intermediate reduced product, which feeds a lower
melting reactor heated with oxygen and powdered coal that produces gas
for the cyclone and the final liquid iron product.
NUCOR/US Steel: a carbide reduction route is another prospect for

increased oxygen usage in smelting; this proposal from US Steel is to use
smelting of iron carbide pellets made near the ore production site (section
4.3.5).
COREX: a full-scale plant is operational and this system is examined in

more detail in the next section.
COREX: pure oxygen blast and direct coal fueling. The key feature of the
COREX process, now in full-scale trials, is an integrated plant tying
together both the iron-smelting units and the power-generation and
oxygen-generation plant. The first major COREX plant was started in
Pretoria, South Africa, in 1990 and now produces over 300 000 tonnes per
annum of molten iron. There are around ten more plants in detailed
planning and it seems likely that this process will have some impact on
annual production by the end of the millenia.
The heart of a COREX plant is the melter-gasifier (Figure 3.3). Here a
pure oxygen blast melts reduced iron pellets and gasifies powdered coal. The
iron is tapped off, along with slag at the bottom, whilst the copious gas
produced at the 1000C operating conditions is taken off to a dust separator
and thence, still hot, to the iron-pellet reduction furnace. Here pellets of ore
(together with slag additives) are reduced from Fe(II) or Fe(III) minerals to
pellets of iron in the hot gas stream. The gas mixture from the melter-gasifier
is designed to be mainly CO and with 30% and provide complete
pyrolysis of all the coal components to these simple gases. Slag-forming
additives such as dolomite and limestone in the reduced iron burden have a
double duty. As well as forming slag with unwanted elements in the ore, they
absorb sulphur from the gasified coal. The resultant sulphuric slag is no more
offensive than standard slag.
The rest of a COREX plant has nothing to do with iron-smelting but is
essential if the process is to be competitive. First, the amount of gas
produced exceeds that necessary for reduction and melting because of the
large amount of coal it is necessary to inject This gas is burnt in a combined
cycle gas turbine/steam turbine power plant, producing 100300 MW spare
electric energy for a typical 3001000 ktonne p.a. plant. The electrical
output is reduced because part of the power produced must be used for the
oxygen plant compressor. Alternative uses may be found for the COREX
gas, depending on neighbouring plants, but a good match must be found to
its output.
3.1.4 Conversion of iron to steel
The conversion of the weak and brittle raw pig-iron into the hard but tough
steels is basically a process of adjusting the alloying elements in the iron.
Carbon must be moved downwards, along with other undesired elements
such as sulphur or particulate inclusions such as unreduced ore, and some
additions must be made.
The basic conversion process is the elimination of most of the excessive
amounts of carbon present in blast furnace iron. This is essentially achieved
by oxidation, today invariably using oxygen bast. The relationship between
the oxygen and carbon content of steels generally is an inverse one, as
expected from consideration of a simple chemical equilibrium
At equilibrium, where k is a constant. Hence, increasing
oxygen levels in the steel by blowing more in will decrease carbon content.
Figure 3.4 Reduction in tramp/alloy elements in steel in BOS process.
The older methods of steel conversion, such as Bessemer converters,
involve blowing air through the bottom of the molten steel vessel (bottom-
blowing) . Air now cannot be used as the iron produced has too high a nitride
content. The modern versions of the Bessemer technique generally involve
an oxygen blow along with a coaxial shield of hydrocarbon such as methane.
Recent converters have been dubbed Q-BOP (quick basic oxygen process).
The endothermic dissociation of the hydrocarbon near the injection pipes
causes a degree of cooling which enables the injectors to survive (with
oxygen in the blow they can be burnt away very quickly) a matter of a few
minutes.
Other undesirable tramp elements, such as manganese, phosphorus and
sulphurs are also removed from molten iron in steel conversion (Figure 3.4).
As well as gases, some additional solid chemical refinements may be
added in the purification process. For example, cerium removes residual
sulphur and oxygen. In addition, there are reactions between the refractory
container and the molten steel; for example, phosphorus is removed by
reaction with basic lining refractories, hence the term BOS (Basic Oxygen
Steelmaking).
3.1.5 Basic oxygen and LindDonawitz/oxygen-lancing Steelmaking
processes
As noted above, oxygen blown through an iron melt will tend to oxidise
sulphur and carbon to gaseous and CO. The LindDonawitz (LD)
process avoids the need for technically difficult holes in the bottom of the
vessel by using an effect of the oxidising process. LD, or oxygen lancing, is
so-called because it is done by lowering a water-cooled lance, a rod fed with
high pressure oxygen, into the Steelmaking vessel. The CO produced in the
Figure 3.5 (a) Bottom-blown converter. (b) LindDonawitz converter.
high-carbon melt results in a stirring action, distributing the added oxygen.
As a result the lance does not need to be immersed in the melt, neither is it
necessary for the jet of to penetrate deeply into the melt. In fact either of
these circumstances would lead to destruction either of the lance or of the
vessel lining.
In practice, the LD process is rapid and efficient; a converter vessel of 200
tonnes or so can be reacted in 20 min or so. The process is also highly
exothermic and the heat output can be used to enhance the process output
by melting added scrap metal up to around 10% of the process. The
produced steel is poured out into a ladle for alloying elements to be added
and then taken off for casting. Variants of the LD process, which is known
generically as basic oxygen steel manufacture after patent battles, are now
the most popular processes in the industry. Variations include additions to
the oxygen blast, for example powdered calcium oxide (lime). This reacts
with excessive phosphorus in the steel melt.
3.1.6 Mini-steelworks and arc furnace steelmaking
There are enormous advantages of scale in blast furnace iron-smelting. For
this reason, small steelworks use scrap steel or, more rarely, prereduced
pellets of iron. Starting from cold metal of fairly low carbon content, there is
no possibility of using the carbon content to raise the vessel temperature, as
in the oxygen injection of the LD process. Melting is therefore carried out
with an arc furnace. Subsequent processing may include vacuum, deoxi-
disers and oxygen/argon blowing. The latter constitutes the argonoxygen
decarburisation (AOD) process (section 3.1.7).
The high cost of electrical heat, and the very large carbon electrodes
needed, make the arc furnace a relatively expensive source of liquid iron. A
number of steelworks now operate arc furnaces with additional fuel burners.
These are switched on early in the steelmaking cycle to supplement the
electrical heat. They use oxygen and coal or oxygen and methane, and have
advantages in terms of spreading heat over the burden more evenly during
the early part of the cycle. As the burden melts, the burners are extinguished
and the arc relied upon to complete the melting process.
Direct reduction. Pellets of sintered iron, produced by direct reduction of
iron ore by hydrogen or methane without melting of the iron, are already an
important feedstock of iron for mini-steelmills.
They have many advantages: they can contain more than 95% iron, with
very low sulphur content. Sulphur is usually removed to a large extent
during the reduction process as volatile or Direct reduction can
also avoid much of the high cost (section 3.1.3) of pelletising ores as it can
operate directly on ores with fines.
3.1.7 Argon for stirring
Molten steel is highly reactive, so much so that the only gases which can be
guaranteed not to react with it are the inert gases such as argon. Argon can
therefore be used for agitating steel melts without the danger of unwanted
reaction. Sparging with argon can thus be used for lowering absorbed levels
of gases such as CO and in melts. For stainless and alloy steels,
chromium, nickel and the other principal alloying elements are also reactive
in melts and argon is useful here, too.
45
The use of argon as an inert blowing
gas is not confined to steels. Non-ferrous metals and alloys which react with
nitrogen, such as aluminium and magnesium, can also be stirred and
degassed by using a blow of argon.
The popularity of low-carbon stainless steels (such as 316L) has meant
that processes for the economical reduction of carbon in bulk stainless steel
melts have had to be developed. The argonoxygen decarburisation process
is now common for the lowering of carbon levels in stainless steels. Blowing
the highly alloyed stainless steel melt with pure oxygen sufficient to reduce
the dissolved carbon down to low levels is likely to oxidise too much of the
alloying chromium and nickel. An obvious solution is to blow with an inert
gas containing only a proportion of oxygen. However, nitrogen would lead
to nitride formation and so is often ruled out. AOD involves bubbling a
45
The reactivity of molten stainless steels is such that for high-grade stainless steels it is now
fashionable to carry out vacuum remelting and sometimes even repeating this process. 316L
stainless steel treated in this way has fewer particulate occlusions, lower dissolved gases and can
be polished/electropolished to mirror qual i t y finish more easily t han regular grades.
mixture of argon and oxygen through the steel alloy melt where the oxygen
combines with the carbon and is carried off as CO in the argon.
The AOD process is the largest scale use so far of argon gas; furnaces used
have output measured in hundreds of tonnes per day. Argon usage of up to
per tonne hot metal processed is reported. Fortunately, argon is
available in large amounts from the liquid air distillation separation plant
used to supply the oxygen to other parts of a steelworks.
3.1.8 Oxygen non-ferrous metal extraction
Many non-ferrous ores are present in nature as sulphides in the reducing
conditions of deep rocks. Many of these, including nickel, cobalt, copper,
zinc and molybdenum, can be successfully extracted by breaking up the rock
and then applying a hot solution of chemicals. Typically ammonia, sodium
hydroxide or, most likely, sulphuric acid, along with air or oxygen, are used
in this hydrometallurgical winning of metal compounds.
For example, is commonly used in copper extraction. Liquors used for
extraction of copper can often be regenerated continuously by bubbling
oxygen through them.
46
Typical reactions are
Oxygen allows a useful degree of process intensification to be attained and
this is the basic justification for its use over straightforward compressed air.
The gas must be pumped into the suspension of mineral and chemicals, and
it is often cheaper to use a small vessel with concentrated ore/chemicals than
a dilute solution. The gas bubbles through, replacing the dissolved oxygen
that has reacted. For a concentrated mix the replacement of oxygen must be
performed quickly. At least five times more air must be pumped than
oxygen, because of the content of air. In practice, much more air than this
is normally required because the bubbles of air become depleted in oxygen
and are then less effective at transferring oxygen to the mix. If a pressure
vessel is employed for the reaction, then, other things being equal, a five
times higher pressure must be used, raising the cost of the vessel and
pipework. Finally, oxygen systems are designed to have minimum gas
46
Another commonly used copper extraction process is that of etching printed circuit boards.
Here the standard process involves the removal of bare copper from between the tracks on a
copper-plated plastic circuit board. The tracks are protected by photosensitive film or lacquer.
The etch solution is usually ferric chloride, which is reduced to ferrous chloride during the etch
However, the activity of the etch bath can be extended by bubbling oxygen through the
solution, which reoxidises the Fe(II) to Fe(III). However, build up of copper and various
reactions and precipitates prevent continuous regeneration.
exiting the reactor, whilst air systems must have most of their flow rate
passing right through, carrying off heat, water and chemical spray. These
side-effects all have a cost and, in practice, oxygen, despite its high cost, can
frequently be justified. This was the conclusion of a number of studies by
Bechtel, Anaconda, Inco, Cominco and others (Denny, 1977).
Oxygen in gold leaching. An unusual hydrometallurgical reaction which
uses oxygen is that of gold leaching. The GOLDOX process, developed in
South Africa by AFROX, uses oxygen to assist in the extraction of gold in
South Africa and elsewhere. The process operates basically as for the
sulphide ores above, with the exception that the starting material is very
poor (only approximately 5 parts per million of gold) and contains tiny
particles of solid gold. The chemicals added are lime and cyanide, and
oxygen is supplied to complete the reaction
The aurocyanide compound is soluble whereas the hydroxide and many
by-products (due to reaction with gangue minerals) are largely precipated.
Oxygen may be added before cyanide addition to oxidise gangue minerals
such as iron suphides, which waste cyanide. The aurocyanide is recovered in
carbon absorber units. The process can use oxygen from the air. However,
with pure oxygen there are advantages in smaller leaching vessels, increased
gold recovery, savings in cyanide and improved carbon absorber lifetime.
Ore roasting. There are problems in dealing with large volumes of
mineral-laden water in the processes described above, however, and hot
smelting of sulphide ores is an old-established and popular process. Just as
oxygen can help in the iron-smelting process it is also used in the smelting of
other metals. For example, using oxygen or enriched air to burn it,
chalcopyrite can be converted largely to impure copper (ready for
electrolytic refining) and sulphur dioxide (useful for sulphuric acid produc-
tion).
Oxygen in non-ferrous smelting again seems a curious anomaly. The use
of oxygen in non-ferrous sulphide smelting revolves around issues such as
the reduction in fuel used to reach high temperatures and the reduction in
waste exhaust gases. The use of oxygen reduces gas clean-up, gives a
reasonably clean source of and reduces all capital equipment size.
3.1.9 Oxygen flame enhancement
Flame emissitivy. The addition of oxygen to a flame increases its adiabatic
temperature, i.e. the theoretical maximum to which it can rise. The high
adiabatic temperature for 100% oxygen addition is very important for pure
oxy-fuel flames in metal cutting and welding (section 3.2.3), while the
addition of oxygen at less than 100% is useful elsewhere in processes such as
iron-smelting (section 3.1.2).
Another example of the use of oxygen addition is in heating tanks of
molten glass. Here addition of oxygen raises the flame temperature. It might
be wondered why this is necessary, given that the melting point of
commercial soda-glass is 900C, well below the adiabatic flame temperature
for air/methane. The answer lies in the inefficiency of heat transfer, much of
which stems from the low emissivity of the flame. A flame transfers heat
from its core to the object being heated by both convection and conduction.
In many situations, where the hot gases from a turbulent flame flow around
the object until they are appreciably cooled, heat convection is important.
However, in other cases heat transfer is mainly via radiation. This is the case
with a glass tank; conduction is limited because the tank is heated by a flame
above the surface of the molten glass and the primary heat transfer
47
is
radiative. The increased flame temperature due to oxygen enrichment leads
to increased radiation, which is large compared to the rise in temperature
(Table 3.1).
The effectiveness of oxygen-enhanced flames can be increased further in
many instances where increased heat transfer is needed by the use of oxygen
lances, separate from the main burner (Monnot, 1985). For example, where a
tank of molten liquid is being heated from above by a flame, an oxygen lance
directing oxygen into the underneath part of the flame just above the liquid
(under-flame lancing) will enhance heat transfer to the liquid whilst not
increasing the general temperature of the furnace top. It also economises on
47
Noting that flame heat transfer is dominated by radiation, some researchers have
attempted to improve the radiation from a flame by means other than increasing its
temperature, i.e. increasing its emmissivity. Noting that it is the infrared-active gases such as
water vapour and carbon dioxide that dominate infrared emission in many flames, some
attempts have been made to add even more infrared-active gases to the flame. Obvious
candidates would be additional fuel gases or vapours such as acetylene, or polyatomics, such as
the freons. Although in principle they would possibly be effective at small concentrations, such
methods have not been shown to be effective.
the use of fuel and oxygen. Local heating enhancements of at least 150C are
achieved in practice.
Oxygen versus regenerative/recuperative burners. Regenerative burners
save fuel by more fully utilising exhaust stack heat from a process and
achieve higher temperatures than the simple adiabatic temperatures
(around 2000C) achieved in the combustion of fuel in air. They do this by
preheating the air, most often with exhaust gases. This can be achieved by
heating the heat exchanger with an auxiliary flame or exhaust (recuperative)
or, more commonly, by heating in a bed of a refractory mineral (regenera-
tive). There should be two or more beds of refractory so that one bed can be
heated whilst another is being used to preheat the combustion air in the
furnace.
This sort of system is most common in large installations such as blast
furnaces or glass works, where the heat regenerators are vast cylinders,
stoves full of honeycomb brickwork. Blast furnace regenerators can
operate up to around 1300C. Regenerative burners are complex, particu-
larly in smaller sizes, and prone to problems but are potentially cheaper to
run, in general, than oxy-fuel burners. Oxy-fuel burners are cheaper to buy
and maintain but need costly oxygen supplies.
Oxy-coal burners. The use of oxygen with coal outside of the steel industry
is still unusual. However, the potential advantages are high: coal, even
micronised coal, is cheap and coal flames have a very high radiant emissivity
and a high adiabatic temperature potential. Oxy-coal is now being tried in
some parts of the world in large steam-raising furnaces.
3.2 Gases in metal fabrication
3.2.1 Gases in metal casting
A surprisingly large proportion of steel and non-ferrous castings are still
produced by pouring molten metal into a mound made out of sand with the
aid of a pattern. The major exceptions to this are relatively small non-ferrous
castings produced in enormous numbers, for example many car parts. Sand
moulds are not simply made like a childs sandcastle from beach sand bound
with sea water, however. It is very difficult to mound sand precisely without
a binding agent.
In the foundry, the sand used is superior in its aggregate make-up to beach
sand. It contains a wider variety of particle sizes, including fine clay-like
material which helps it to bind together and form smoother surfaces, and is
carefully adjusted for moisture content. For many purposes this is sufficient.
For more predictable properties, more robust moulds and better casting
finish, however, much sand-casting uses binders either in the bulk sand or
just on the mould surface. These binders are typically organic materials.
Starch is one simple possibility for stiffening up the sand, although higher
performance is obtained from resins, which can be induced to set hard, at
least on the surface, by heat (such as phenolic resins), a catalyst sprayed onto
the surface (epoxies can be sprayed with organic peroxides to achieve this)
or by a reaction with a gas. In the latter category are silicate binders which
set with the application of carbon dioxide. The surface-hardening binders
can withstand the impact of molten metal, which can simply wash away weak
sand surfaces.
Sand used for moulding is largely reclaimed and used again. Here organic
binding resins present problems as a proportion of the sand used must
always be dumped and this disposal is a problem. Another problem is the
fumes emitted by the sand during the casting process; organic resins evolve
unpleasant fumes. Finally, the bound sand must be removed from the
casting and this can be a problem for sand stuck in the recesses of a casting if
there is an appreciable amount of binder present. One process (EFFSET
)
has been developed which neatly sidesteps all these problems and provides
at once higher quality and environmental friendliness: ice binding of sand
moulds. This involves preparing the usual sort of sand/clay/moisture
mixture, shaping the sand around the master pattern by the usual ramming
process, then removing the pattern and freezing the sand with a spray of
liquid nitrogen. Once the surface is frozen the mound is very robust, being
an aggregate/ice mixture. It is mainly aggregate and at the opposite end of
the scale from the impressively strong pykrete (see section 3.7.2), which is
mainly ice.
Surface oxidation and dross formation in casting can be reduced if an
inerted path can be arranged from the melting ladle to the mould. Gas
inerting of moulds is possible, as is the use of inerted enclosed ladling
facilities. Even when only the mould is inerted, quality improvements are
seen. Argon gas is a good choice for this because is completely inert and
because, being heavier than air, it can be used to fill up mould spaces open at
the top (left there to allow liquid metal in and for overflow), rather like
topping up with water. Argon inerting is commonly used on both sand-
casting and on die-casting installations.
3.2.2 Furnace atmospheres for metal processing and heat treatment
Heat treatment of metal parts has in the past often been carried out by
heating parts coated with a substance which will react at furnace tempera-
tures in a suitable way but the coating process can be laborious and the parts
may be disfigured by oxidation in places due to air in the furnace. Even more
laboriously, carburising used to be commonly achieved by packing parts into
carbon inside a furnace. Today, a suitable furnace atmosphere can achieve
heat treatment without oxidation or other undesired surface reaction, and
the gaseous nature of the treatment allows all part surfaces to be processed.
A mixture of inert and reactive gases, either oxidising or reducing, is
employed. Mixtures based on a preponderance of nitrogen are inexpensive
and often effective. Nitrogen is nowadays most often supplied in pure form
by gas companies, although inert gas generators (section 2.2.1) may be
employed. Argon may be necessary as the inert gas for highly reactive
metals, although it is much more costly and has a competitor in the form of
vacuum processing.
Heat hardening or softening in inert atmospheres. The use of an inert
atmosphere in handling metals is of general importance because of the
reactivity of air (mostly the oxygen and moisture in air) with most metals.
Even with metals which are unreactive at room temperature and modest
heat, heating of metals in air to the 500C or more needed for metal working
mostly results in tarnishing due to oxide film. It is often reasonable to accept
the tarnishing or remove it subsequently by polishing. However, polishing
can be expensive, needing additional machinery, and may be labour
intensive. It is, therefore, very useful to be able to replace the air in a furnace
with another, more inert, atmosphere.
Annealing. Bright annealing is carried out in furnace atmospheres con-
taining hydrogen. The result is a clean surface finish (no oxidation) and, with
the correct temperatures and low cooling rates, typically a soft condition,
ideal for most metal-working. Pipes in stainless steel and copper are bright
annealed as part of their processing, between cold-drawing processes, to
make them easy to bend.
Annealing is carried out on metals that have undergone any kind of
cold-working. Hardening occurs when a metal undergoes a large plastic
deformation as in cold-washing. It is often undesirable because, for
example, it may make further cold-working more difficult or may allow
cracking in use. The component cracking occurs because non work-
hardened regions of the component will yield more easily, being softer,
whilst the work-hardened regions will be forced to take more stress and may
crack. The work-hardening itself represents energy stored in microscopic
prestresses, which are relieved when heated, as the metal flows to eliminate
the strains.
Annealing of magnetic iron laminations used for the core in transformers
is a particularly important application of gas atmosphere annealing. Iron
laminations, rather than bulk iron, are used in transformers to avoid eddy
current losses in the core. The laminations are often annealed in reducing
furnace atmospheres, although the object here is not to retain a bright
surface; the laminations will be hidden from view and protected with varnish
in service. (Silicon/iron alloys, with suitable treatment, can also form their
own glassy insulating oxide surface which does not require varnish.)
Annealing heat treatment of iron laminations improves their permeability
and reduces their hysteresis, making them more soft in a magnetic sense.
Laminations are often made from cold-rolled low-carbon iron sheet and
then further cold-worked, in effect, by the stamping process used to shape
them. Other laminations are made from 3% silicon/iron alloy, also low
carbon, which has a permeability approximately ten times higher than that
of simple iron. The deformations result in hysteresis, which leads to losses in
the transformer application directly since the heating of the transformer iron
core is proportional to the hysteresis loop area. Deformations also reduce
magnetic permeability, which means that there are higher Joule (resistive)
heating losses in the conductor coils. Annealing treatment is therefore
doubly valuable.
Decarburising. Decarburisation is also useful for transformer iron lami-
nations. An iron with the lowest possible carbon content is desired because
then the maximum proportion of the highly magnetic ferrite (soft
magnetic) iron phase is obtained with no pearlite. Pearlite is a phase of
steel, comprising microscopically closely spaced layers of cementite
iron carbide, a hard, non-magnetic compound) and ferrite. This phase is
formed when 0.9% carbon is dissolved in molten iron. Even with 0.1%
carbon, a transformer lamination might be 10% undesirable pearlite. The
presence of the other phases also prevents the remaining ferrite from being
so effective. Gas decarburisation using a mixture of nitrogen (to reduce
cost), hydrogen and steam, or possibly a nitrogen/methanol/steam mixture,
is a possible solution to further reduce carbon content. With 80 or 90% of the
gas being more-or-less inert nitrogen and only 1020% of the more
expensive active gas, the following reaction is principally responsible for
decarburising effects
By this reaction it is easy to decarburise typically thin transformer
laminations (approximately 0.5 mm) and it is possible to decarburise up to 3
or 5 mm laminations with longer anneal/decarburise times. The water/
hydrogen ratio for the optimum decarburisation rate is reported to be in the
range 0.20.5. The hydrogen presumably prevents damage to the steel by
restraining the oxidation possibilities of steam on red-hot iron
A temperature of approximately 800C is necessary for this reaction, with
heating for a few hours followed by cooling rates in anneal of as low as 50
per hour (faster for thinner laminations). A dwell time in the iron
temperature of about 450C in an oxidising atmosphere leads to a thin oxide
layer, which in certain thicknesses gives a blueish colour. This part of the
process is therefore often called blueing.
Figure 3.6 Hardness traverses (top) and residual stress distributions (bottom) through the
surfaces of: A, nitrided 722M24 (80 h); B, carburised alloy steel and C, induction hardened
Cr-Ni-Mo steel (Harper and Parrish, 1985).
Carburising. If many steel grades, including low carbon varieties, are
heated in the presence of carbon or carbon-containing compounds to 1000C
or so, then hard martensitic steel and hard carbides will form in the surface
and for up to 2 or 3 mm into the bulk of the material (depending upon time
and temperature of treatment). As for nitriding, the carburised components
can retain a tough core. Originally carburising simply involved packing steel
parts into a charcoal mixture and then heating. Nowadays gas processing is
used as it is more controllable and leaves the parts clean (Figure 3.6).
The carburising mixtures can be a blend of nitrogen and a carbonaceous
gas such as propane (or methanol vapour), usually with a little adventitious
air in the typical industrial furnace. If the gas exiting the furnace is
flammable, it is normally burnt off using an automatic ignitor to minimise
explosion risks; treated steel parts come out in an impressive gush of flame in
the more old-fashioned type of furnace. Inside the furnace the gas is too rich
or too diluted with inerts to catch fire. Fuel gases generally crack at the
temperature of operation into CO and water vapour. Carbon dioxide
can be used in carburising mixtures.
The carbon potential (the percentage of carbon that will be incorporated
in the steel surface) of the gas mix must be controlled. Carbon potential
depends not only on the concentration of C in the furnace gas, but also on
the percentage of water vapour present; more water vapour leads to a lower
carbon potential. For example, a gas mix that reacts to produce 20% CO and
40% (from methanol) in would vary from 0.9% carbon potential at
0.3% water to 0.2% carbon potential at 1.2% water. An Ellingham diagram
is sometimes used to predict what oxidation/reduction effects and carbon
potential a particular gas mixture will achieve at operational temperature.
This diagram is essentially a nomogram recording equilibrium reaction
constants versus temperature.
Nitriding. Nitriding is a lower temperature process (approximately 500C)
than carburising. Originally carried out by immersing steel parts in molten
salt baths based on sodium cyanide, the process is nowadays mostly based on
ammonia as the active gas. The ammonia cracks to some extent on the
surface of the steel, thus incorporating atomic N into it. The nitriding
effectively inserts into the zone near the steel surface minute particles of
nitride compounds, which strain the iron lattice and render the steel harder.
The layer resists corrosion but penetrates only 1 mm or less into the bulk of
the metal. Thus the steel retains its toughness qualities and is resistant to
shattering, whilst having a glass-like surface hardness.
Argon and vacuum atmospheres. Vacuum is an alternative to furnace gas
inerting with argon. Aluminium, magnesium, titanium, etc., can all be
processed under argon. It is interesting to consider the costs involved in
these two alternatives:
vacuum solution: vacuum pumps: high capital costs
strong vessel: high capital costs
pump-down time needed slows down process
argon inerted solution: thin-walled vessel: low capital costs
argon supplies: high running costs
purging time needed slows down process but
this is a less serious defect because capital costs
are low.
Clearly, if the emphasis is on mi ni mum capital costs, then argon inerting is
indicated, and vice versa. Vacuum may be most appropriate for sites with
high utilisation factors where capital costs are less important.
Endothermic generators versus nitrogen-based atmospheres. Where a
nitrogen-based furnace atmosphere is proposed, an alternative to simply
assembling the components of the atmosphere from bought-in gases is to
employ an inert gas generator to produce a nitrogen-rich atmosphere. A
more sophisticated version of the inert gas generator described in section
2.2.1, an endothermic generator, burns fuel with air but restricts the oxygen
supply to give a CO and content in the nitrogen-rich mixture. CO/H
2
content in a nitrogen-based mixture is normally achieved by adding
methanol to nitrogen. The heat of the furnace causes cracking to CO and
The reaction needs the supply of external heat and yields, for example,
20%/40%/40% if methane is used as the feedstock.
The use of relatively pure fuel supplies makes the control and output gas
quality of an endothermic gas generator more acceptable than would
otherwise be the case. With these higher purity feedstocks, a hot nickel
catalyst bed can be used, which is heated externally. Residual fuel, nitrogen
oxides, CO and sulphur compounds will all, however, be present in trace
quantities and may cause process problems or corrode the pipework. Like
inert gas generators, endothermic gas generators are only nowadays used in
processes with minimal requirements on gas quality and a steady demand
pattern, and then only when capital and running requirements must be
minimised and the size of gas stream needed is not too large. The use of
cryogenic nitrogen, either on-site or in bulk liquid tanks, is most often
indicated. Finally, the very high CO and content means that the
flammability problems of endothermic atmospheres are very serious:
equivalent production is often possible with atmospheres which contain
more than 90% and is much safer (carburising operations generally need
the high CO content, however). Nitrogen availability also provides safety
benefits in being available for purging of atmospheres prior to access;
endothermic generator furnaces are at risk from fire or explosion during
shut-down and start-up operations.
3.2.3 Flame welding and cutting
Flames. As noted in section 2.5.4 a flame is the reaction zone of a fuel or
fuel/oxidant mixture with an oxidant. Most industrial burners use premixing
of oxidant with fuel but with insufficient oxidant for complete combustion.
The flame structure then comprises two parts: the central reaction zone of
the premixed gases and a surrounding reaction zone where further oxidant is
allowed to diffuse in and complete the combustion. An exception to this is
the pure fuel diffusion flame used for carbon-coating using acetylene. By
contrast to the relatively dim blue optical emission of the premix flame, the
diffusion flame for any hydrocarbon gas is a highly luminous yellow.
Although normally in dynamic equilibrium, a burner flame can move
upstream (lift-off) or, for premixed flames, downstream (lightback or
blow-back). Generally, a burner that runs satisfactorily on a certain
pressure of a certain gas can be run satisfactorily on a range of different gases
after adjustment of pressure or replacement of orifices. However, if it is
desired to run a burner on a different gas without adjustment, then the
replacement gas must be checked for its Wobbe index or Wobbe number.
The Wobbe number is the calorific value per unit volume divided by the
square root of gas density. Gases of the same Wobbe number can usually be
used at the same pressure on the same burner without adjustment and
without changing the heat output of the flame. This is because the flowrate,
Q, of gas, density, Wobbe, W, and calorific value, C, are usually
determined by a restriction of some kind, which will follow an orifice
plate-type of pressure drop equation
The replacement gas will have the same pressure P, as the original gas,
either because its supply pressure is the same or because pressure regulators
in the system and at the burner are not changed. The flow rate Q', of a
replacement gas of density Wobbe W' and calorific value C will
therefore be given by
Hence
Note that the units in which the Wobbe number is expressed are irrelevant
since the heat output change seen on changing gas to a burner is given by the
ratio of Wobbe numbers. Note also, however, that even a gas of correct
Wobbe index will not guarantee that lift-off or blow-back will not occur on
changing gases, since the new gas may differ in other properties such as
flame speed from the original gas.
Oxy/acetylene cutting. With metal thermal-cutting processes, the metal
must typically first be heated to beyond its melting point over a small
localised area. The molten metal must then be ejected somehow. In flame
cutting, the metal is generally burnt by using an excess of oxygen in the
flame. The oxidation of the metal is itself exothermic and generates
additional heat, making the process a highly efficient one.
Ingenious designs of burners are employed to try to optimise the cutting
speed and efficiency. The most common designs involve a mixture of oxygen
and acetylene issuing from an annulus for heating the metal with a nozzle in
the centre for cutting oxygen, the oxygen flow being switched on and off
conveniently with a trigger on the torch (Figure 3.7). A metal mesh filter is
connected in the gas supply lines to prevent flashback, i.e. the flame from the
burner tries to travel back up the flexible supply hoses.
Figure 3.7 Oxy/acetylene cutting torch for cutting thick steel slabs.
The oxy/acetylene cutting process is, perhaps surprisingly, effective
having:
very high efficiency in the use of the relatively expensive acetylene fuel,
although copious oxygen is still needed;
very high speed, even on thick sections; sections up to 2 m or more are

possible, whilst speed is such as to match steelrolling-mill output on
specially designed high speed cutting burners;
very narrow kerf width, down to 1 mm or so;
a surprisingly neat surface finish with striations often below 0.5 mm along
the cut edge.
There are many specialised designs of cutter, those used for scarfing billets
in steelworks, for example, or gouging nozzles for making channels in thick
steel sections. It is even possible to inject extra oxygen via a separate lance at
ultrahigh pressure, thereby burning and ejecting steel even faster. A patented
device from the Messer Griesheim Corporation (Germany) employs a liquid
oxygen feed right to the burner head at high pressure.
Steel plates which have been cut with oxy/acetylene do have a hardened
region or heat-affected zone (HAZ) which can make subsequent machining
difficult. The hardening is due to the formation of an interstitial iron/carbon
phase in the steel (martensite). This forms when the steel has cooled too
quickly to allow the formation of stable, tougher iron/carbon phases such as
pearlite. However, this hardened zone is acceptable in many mild and other
steels, and can be ground out if necessary. If the next operation to be
performed is welding, then the hardness of the cut zone is not such an
important factor as the welding operation will introduce a HAZ of its own.
The HAZ is actually used in the process of flame-hardening, where an
oxy/acetylene flame is used to rapidly heat up the surface of a steel
component, whilst immediately after heating the steel is cooled with a jet of
water. Special burners incorporating a water-cooling jet just behind the
flame are often used. Because of the rapid heating, the depth of metal
hardened is limited, leaving the core of the component still tough.
Oxygen purity effects. Oxygen purity is now so high (99.9%) that effects
are not normally seen. However, in cutting of steel in the past, when oxygen
purities from 97% upwards were commonly supplied, cut rates and finish
quality were found to be related to the oxygen purity. The Welding Institute
in Cambridge accumulated much data on this, as did shipyards in the former
East Germany.
More recently work with laser cutting has shown that oxygen purity effects
can be seen in the quality of cut in the more controlled conditions of a
laser/oxygen cut. The effect of (typically argon nitrogen) inert impurities
would appear to be the formation of an insulating blanket of gas, slightly
reducing heating and reaction of the molten steel with oxygen. The blanket
forms as the molten metal reacts instantly with the oxygen, leaving the inert
gas fleetingly in a thin layer above the kerf.
Oxygen cutting processes in steelworks. Oxy-fuel cutting is a very large user
of oxygen in the steel industry. Scrap must be cut up so that it is small enough
to go in the furnaces. Large ingots will typically need cutting into smaller
ones; sometimes this can be done whilst the ingots are still at red heat,
sometimes on cold material. Ingots are often split up into different sections.
Steel from the middle of an ingot, for example, has different properties from
that at the edges and different end-applications will need those differing
properties. Sheet material is often cut by multiple oxy/acetylene cutting
torches these can be arranged to cut diagonally across the metal so that a
cut at right angles can be made on a continuously moving sheet.
Scarfing, a specialised form of cutting, is used to remove surface defects
from ingots prior to rolling and further operations, and is a significant
oxygen application. Automatic scarfing equipment blasts high-pressure
oxygen, together with propane fuel, at billets of steel as they pass at a
walking pace, removing a millimetre or two of surface in a spectacular
shower of sparks. In this way high surface quality can be guaranteed on steel
plates and coils. This is particularly important in thin sheet.
Alternative oxidants and fuels for hot flames. The nitrous oxide/acetylene
flame has been suggested and used on occasions because nitrous oxide can
be stored as a liquid in relatively lightweight high-pressure cylinders,
whereas oxygen must be stored as a gas in heavier higher pressure cylinders.
This means that an flame kit can be made to run for longer
from the same size and weight of gas bottles than an kit, but still
has a very hot flame.
Nitrous oxide flames are also employed in some flame spectrometers. The
flame vaporises aqueous samples, typically mixtures of trace metals, to
temperatures where they emit light in characteristic line spectra. Analysis of
the spectrum emitted gives the concentration of each species down to parts
per billion or lower concentrations. emissions are likely to be higher in
nitrous oxide-based burners. (In coal-firing it is well-known that nitrogenous
compounds in the coal fuel lead to higher levels in the exhaust gases.)
The temperature achieved in an oxy/acetylene flame is the highest of the
common practical flame systems. A list of a few theoretical adiabatic flame
temperatures gives a flavour for the maximum achievable (Table 3.2).
The figure in the last column of Table 3.2 raises the temperature of the flame
to the fourth power and normalises on air/propane to give an idea of the power
radiated by the flame if it follows the StefanBoltzmann radiation law.
The use of a vaporisable liquid fuel gas has clear advantages in terms of
cylinder calorific capacity. The MAPP (methylacetylene/propadiene)
mixture is popular in the USA, rather than acetylene, because it provides a
very hot flame at lower cost and more conveniently. Propylene is also
favoured in applications such as flame-spraying where a large gas capacity and
usage is expected.
More exotic flames are also possible, as shown in Table 3.2. Cyanogen is a
hazardous gas that is only manufactured and used as a laboratory curiosity. It
is poisonous and condenses at 21C. Fluorine is too dangerous, as is its
combustion product (HF), to be considered as anything other than a curiosity
with regard to flame heating.
The achieved adiabatic flame temperatures listed in Table 3.2 are functions
not just of the released heat of reaction but also of the heat capacity of the
combustion products and the degree to which the final products are
dissociated. With air the flame is cooled because of the requirement to heat up
the inert nitrogen. Similarly, dissociation of the reaction products tends to
reduce the flame temperature of hotter flames. There is also an effect that the
measured temperatures are reduced by the production of free radicals and
incompletely reacted products in the flames. This can be easily verified by
measuring the electrical resistance of a volume of gas at modest voltage
(normally infinite) and repeating the experiment with a flame. Most flames
show a surprisingly high conductivity, though still in the gigaOhm or more
region, showing the presence of ions and free radicals. An acetylene flame, for
example, is a potent source of free radicals.
The order of flame temperatures is not the correct parameter for ranking
fuels for a vehicle, for example, where heat of reaction is important, and
different again from the ranking for rockets where heats of reaction per unit
weight and molecular weight come into the equation.
The relative effectiveness of these high temperature flames in the task for
which they are most usually required, heating an already hot solid hotter still,
is a very steep function of their temperature but also depends on their
emissivity. The StefanBoltzmann law for thermal radiation of grey bodies
illustrates the effects. It states that the power, P, radiated by a bodys surface
is related to a constant for the surface the emissivity, and the absolute
temperature, T as follows
The complete problem is, in fact, more complex than this owing to the fact that
thermal emission from flames does not follow the simple Planck spectrum,
which is a smooth broad peaked distribution of power. The frequency of the
peak varies with temperature: white when hot, yellow, orange or red when
cooler. Instead, the infrared emission is banded, with polyatomic species
emitting much more effectively than diatomics or monatomics (section
2.3.5). This may account for some of the claimed advantages of the bubbler
systems sold for flame enhancement. These add a vapour from a bubbler filled
with liquid, consisting of highly unsaturated hydrocarbons, to the basic fuel
gas stream. An enhancement of the flame temperature and effectiveness is
claimed for the systems in the case of the lower performance flames such as
oxy-methane.
In the radiation of visible light the particulate carbon content of the flame
seems to be the most important factor. This explains why burning hydrogen is
almost invisible.
48
The hydrogen flame emits light and becomes electrically
conductive in exact proportion to the rate at which carbon atoms are added to
it. (Advantage is taken of the latter effect in gas chromatograph instruments
with flame ionisation detectors as mentioned in section 2.3.4.)
48
The near invisibility of hydrogen flames is a real potential hazard. Hydrogen flames can occur at
leaks and not be perceived until the heating effect of the flame has disastrously melted or set fire to
something else. Hydrogen does show up in the most surprising places and, if ignited, can cause
trouble. Caution is advised even near simple pieces of domestic apparatus such as a battery whilst
it is being charged or a hot-water radiator whilst it is being bled to remove trapped gases. In both
these cases small quantities of H
2
can accumulate and, because of the minute ignition energy and
wide flammability limits of hydrogen, start fires.
Thermic lancing. An unusual example of a fuel oxygen flame used for
cutting is the thermic lance. The thermic lance is one of the few devices that
will cut through rock and concrete, even steel-reinforced concrete. The lance
consists typically of a bundle of steel rods inside a steel tube of approximately
20 mm diameter. Oxygen under a few bars pressure is passed down the tube
and the end is lit with the aid of an oxy/acetylene torch. The result is a
spectacular white-hot firework flame with formidable penetrating powers.
Note that once lit the thermic lance is running on solid fuel, i.e. the steel rods
and tube.
The thermic lance is now falling out of favour, however, owing to advances
in the art of diamond-tipped drilling for concrete; the drill makes a much
neater though much slower job than the lance. However, a typical 3 m lancing
rod, driven by a normal oxygen cylinder, can cut a 3 cm hole in reinforced
concrete 50 cm thick in approximately 3 min. The noise, copious clouds of
smoke and blast furnace heat of the lance are drawbacks, however.
It is possible that thermic lancing on a smaller scale could be more attractive
and small (6 and 10mm diameter) lances which are electrically ignited are now
available. They can be started by earthing one end, passing oxygen over them
and igniting with the small arc on a copper plate connected to a car battery.
3.2.4 Flame spraying/coating
Hard-facing with welding rods. It was noted early in the history of welding
that the weld surface on mild steel was much harder than the parent metal
(section 3.2.3). This led eventually to hard-facing softer metals with harder
metals and from there to the whole industry of flame-coating and
flame-spraying. In its simplest form, a coating of, say, hard Stellite alloy can
be applied as a welding rod to a mild steel or other substrate exactly as if a
welding operation were being carried out. Subsequently, the workpiece may
need heating to consolidate and fuse the deposit, and the surface can be
ground completely flat if necessary. This sort of operation has often been used
as a maintenance tool, with a worn component being built back up to its
original size and form with the weld metal. The repaired article, being
hardened on its surface, so slowing the wearing process, will often improve on
the lifetime of the original monolithic article.
Flame-sprayed powders for coatings. This technique has now been
extended to include much more precisely controlled processes with a flame
held a precise distance from the substrate (usually using an automatic
machine) and actually keeping the substrate out of the direct flame. The
coating material is added as a very fine powder to the flame gases. The red-hot
minute particles of coating impact on the substrate surface and partially fuse
there but do not heat the substrate enough to damage it, which is the case with
welded-on hard-facings. The flame-spraying technique using powders is also
more versatile because the substrate is not unduly heated and the coating
applied can be very thin but just as effective. The thinness of the coating is
Figure 3.8 (a) Thermic lancing of masonry. Note stream of molten iron oxide and minerals
running out of hole. (b) A Sulzer Metco plasma spraying system being used to spray a component
of an aeroengine. The spray gun is mounted on a robotic arm which moves in five axes, the robot
itself being moved linearly to provide a sixth axis (courtesy of EMMS).
often an advantage because no further grinding or machining operation need
be performed.
A large number of materials can be flame-sprayed, not just metals.
Ceramics such as carbides or alumina (for wear resistance), calcium minerals
(for rendering metal screws and parts biocompatible) or cermets (metal/
ceramic composites) can be treated. Patents covering boron nitride,
zirconium nitride, chromium carbide in nickel and refractory silicides have
been filed, covering applications from simple wear resistance to protection of
metals from attack by other metals in the molten state.
Increasingly sophisticated apparatus is now used for flame-spraying, in
which high-pressure fuel gas and oxygen are fed to a supersonic nozzle, or a
detonation gun (the Praxair patent process), using fuel/air mixtures (Figure
3.8). The flame nozzles are often designed to use a very high flow rate of fuel
and oxygen at a relatively high pressure (up to approximately 10 bar). A
liquefied high-energy gas such as propylene or MAPP is often specified
because the quantities needed are inconvenient with the much less efficiently
stored acetylene and because of pressure limits on the latter. However, with
draw-off rates of many cubic metres per hour, it is often necessary to use a
liquid withdrawal from the cylinders of fuel and a heated vaporiser unit.
The metal or ceramic powder to be coated achieves quite high temperatures
and very high speed (up to 1000 Despite these extreme conditions, the
substrate must not be heated unduly in order to avoid damage and differential
thermal expansion problems when the coated parts cool. The maximum
substrate temperature depends on substrate and coating but is often a
rate-limiting step in the coating process as temperatures as low as 100C are
sometimes required. Despite very high fuel/oxygen powers (100 kW is not
unusual), therefore, the coating rates are limited: (a recent paper quoted
20 m per minute for coating a commonly required tungsten carbide/cobalt
composition).
Coating rates can be improved by using forced air cooling on the substrate
immediately after a layer has been coated. In other cases, liquid carbon
dioxide has been sprayed onto the substrate immediately after coating.
Liquid nitrogen is also being tried. When coating rotating substrates, cooling
can be achieved by simply following the coating flame gun with a cooling blast.
When coating flat substrates effectively, instant cooling can be achieved by
raster scanning the flame gun with an adjacent cooling blast gun and thus high
coating rates can be achieved.
Flame welding. Welding with an oxy/acetylene torch (flame or gas
welding) is still a very popular process. It is not, however, now regarded as
giving the highest quality joint for many purposes. The weld produced by a gas
torch is often inferior to that produced by an inerted arc process. By careful
choice of welding rod, flux and practice of technique, however, perfectly
satisfactory results can be obtained.
The technique of gas welding is perhaps more of a manual skill than arc
welding (section 3.2.5). The problem on thinner sections is that the
welding torch wants to be a cutting torch and beginners often burn right
through the workpiece. This problem can be restrained by small circular or
side-to-side motions of the torch but it requires skill to achieve good
penetration of the weld at the same time. In an ideal situation, the metal
plates to be welded are prepared with a V channel between them of a
little under 90. The torch is angled up the channel and used to produce a
small pool of molten steel into which the filler rod, angled down the
channel, is melted. The whole process is then moved steadily up the joint.
3.2.5 Arc welding and cutting
With the development of high-current power supplies grew the realisation
that extremely high temperatures, far higher than flames, could be reached
by striking and maintaining an arc. The first processes used an arc struck in
air between the workpiece and a rod of weld metal. Because of oxidation
and nitrogen embrittlement caused by the high oxygen content of the arc,
flux materials were packed around the welding rod, resulting in the familiar
stick electrode process used today for rough welding work. The use of
shielding gases to substantially keep out the air was a later development
that enables much higher quality welds to be obtained. A landmark in the
development of arc welding was the large-scale promotion, beginning in
1948, of the MIG (metal inert gas) concept by the Air Reduction
Corporation (now AIRCO) in the USA.
Plasma and arc cutting. Plasma cutting involves transferring a powerful
stream of highly ionised gas, also containing many atoms and molecules in
an excited energy state, onto an earthed metal plate. There the highly
energetic neutralisation and deexcitation processes of the stream rapidly
heat the earthed plate to melting point and the high-speed stream then
carries the melted metal away leaving a relatively neat cut. The arc is
struck between a tungsten electrode and a water-cooled copper nozzle, the
whole torch being maintained about 10 mm from the workpiece. Arc
cutting employs a simpler torch in which the arc struck is between the
tungsten and the workpiece. Used only for rough work, arc air cutting
relies on blowing air at pressure through an arc struck on a metallic
workpiece.
Argonnitrogen and argonhydrogen mixtures are used for high-quality
cutting, giving kerf widths of about 3 mm in a variety of materials including
aluminium, steel, copper and other non-ferrous materials. The fact that the
method does not rely on the oxidation of the material, as in oxy/acetylene
cutting, gives it greater versatility. There are some safety implications for
plasma cutting: the exhaust gases contain and and must be
extracted, whilst noise levels are high.
Argon arc welding. Arc welding can be performed with stick electrodes
which are coated with a thick layer of flux materials, such as carbonates,
which evolve thus providing a gaseous shield, or with fluxes which form
a glassy slag film that isolates most of a joint from the air. However, manual
metal arc welding, with stick electrodes, does not in general produce very
high quality joints. The arc is difficult to stabilise, the glassy film of slag
makes the finished joint messy, whilst air still penetrates the slag and causes
oxide and nitride formation making joints in steel brittle.
Inert gas arc welding employs two properties of argon: the stable electric
arc it forms at a few hundred amps and a few tens of volts, and its inability to
react with metals at white heat. The concept of the process is very simple. A
few hundred litres per hour of argon-based gas are passed up a tube around
an electrode to which a suitable power supply is connected. An arc is struck
on the workpiece and heats it while the gas flow forms a blanket, preventing
excessive oxidation.
Careful control of the process is essential and the following points must be
noted:
metal must not drip out of the joint;
the heat-affected zone must be minimised and have acceptable metallur-

gical properties;
inclusions of oxides or nitrides will lead to weakness;

segregation of alloy components is usually undesirable;
excessive oxidation must be avoided;

the weld material and HAZ should retain similar electrochemical
properties to the parent metals in order to avoid joint corrosion;
blobs of molten metals ejected from the joint area, spatter, is difficult to
clean off afterwards and should be controlled to a low level.
Naming of arc welding processes. The names of electric welding processes
are currently going through a transitional period in which new names (which
more accurately describe the processes) are supplanting the older terms
(Table 3.3).
Tungsteninert gas (TIG) welding. An argon arc can be used simply to
transfer heat to the parent metal. By ensuring a close fit between the parts
being joined (a fit of a fraction of a millimetre is possible for many small
components and for larger components with regular shapes) all that is
needed is for the arc to provide the necessary localised heating in an inert
atmosphere to melt both metal edges and fuse them together. Addition of
helium to the argon is used for al umi ni um, whilst hydrogen is added for
stainless steels. For duplex stainless steel argon with a small (less than 3%)
amount of nitrogen can be used. (Duplex stainless steels are a newer series
of alloys which have a two-phase structure and offer superior strength and
corrosion resistance compared to standard stainless steel.)
The TIG welding torch is a simple device with a tungsten tip and a
ceramic nozzle to protect the end of the gas supply tube. The tungsten tip is
usually made from tungsten with additions of thorium or zirconium oxides,
which increase electron emission and ease starting of the arc. Depending on
power level, the tip can be anything from 0.5 to 6 mm or more in diameter.
The power supply is usually a versatile device capable of supplying positive,
negative or a.c. voltages up to 100 V or so. Aluminium is usually welded with
an a.c. electric supply, the positive half-cycle allowing positive ion
bombardment of the tenacious oxide film on Al, which inhibits its welding
with a d.c. supply. Most other metals are joined with a negative electrode
d.c. supply to reduce loss of tungsten. Pulsed power supplies are now
common in professional equipment as well as arc-striker high-voltage pulse
or radio frequency (RF) burst options. The power supplies are nearly always
arranged to drop in voltage in a controlled way as current increases; this
stabilises the arc produced.
The whole process is fairly easy to automate and small inexpensive
automatic TIG welders are now common. Simple geometry welds, such as
pipe butt welds, are now commonly performed only by automatic orbital
pipe welders. These are clamped onto the two pipes to hold them butted
together. Inside the weld head are electric motors to rotate (orbit) the head
around the pipe. These can be programmed to allow for differences in
welding at different ambient temperatures, orientations and with different
metals. Once programmed, they produce virtually perfect welds consist-
ently with very little fuss.
Metalinert gas (MIG) welding. MIG welding replaces the tungsten
electrode with a piece of filler wire that is automatically fed by an electric
drive up to the torch head. It is now a very popular process in industry and is
commonly seen in the motor trade where an oxy/acetylene burner would
previously have been used. It can be carried out at high speed on thick
sections but is also very good on the thin plate used in vehicles. Gasless
MIG welding relies on a barium carbonate-based flux in the middle of the
feed wire to produce gas (although it also evolves an unhealthy amount
of barium). Gasless MIG welding suffers from quality problems similar to
those of stick-electrode welding, however, and it is difficult to make the wire
with enough flux to be effective and still leave it easily handleable.
MIG welding does not work well with pure argon. Early MIG processes
relied on however, is today only used occasionally on mild steel,
mainly in low-cost work, because it is more difficult to control and has
increased spatter relative to the new gas mixtures. The success of most MIG
welding depends on a mixture based on argon or helium. Oxygen at 1% or so
produces a more stable arc and better fusion and penetration via lower
surface tension in the weld pool. Increasing helium content stabilises and
increases the arc heat output. addition in the mixtures
improves wetting in the weld relative to argon and results in very low
spatter. Argon plus 5% or similar is now the most common welding gas
mixture for MIG processes.
The rate at which the filler wire is fed to the welding head determines to a
large extent the welding current (which may be anything from 100A to
600A) and the rate of welding. The filler wire is copper-plated to avoid
corrosion and possible jamming in the head feed. Depending on the arc
current, arc to work distance and other parameters, metal may be
transferred from the consumable electrode to the work in three distinct
ways:
at short arc distances by the electrode actually dipping into the molten
weld pool;
in longer arcs by high-speed formation and transfer of droplets a few

millimetres or less in diameter;
in longer arcs by a fine spray of liquid metal from the electrode.

The voltage/current characteristic of the arc depends on the metal transfer
mode. In dip-transfer large currents pass at low voltage, whilst in spray-
transfer currents are lower and voltage higher. The gas type affects the
process, with argon-based gases encouraging much higher metal transfer
rates.
The MIG welding torch is slightly more complex than that used in TIG
welding. The feed wire goes down the centre of the tubular central ceramic
nozzle and is surrounded by a gas supply tube. A switch on the torch starts
and stops the wire feed to the unit. Wire from a motorised reel mounted on
the power supply is pushed up the torch tubing, which also provides gas,
electric current and, on large units, water cooling. The torch may also need
to have a puller unit for feed wire if the wire is too thin to be fed from the
power supply end.
Noxious welding fumes. The role of gas in the production of noxious
welding fumes and gases has been studied in recent years. Although fumes
and gases from arc welding should be sucked away if at all possible by an
exhaust hood, in practice it is still an advantage for the operator if less
noxious exhaust is produced. Using lower currents for welding often reduces
fume production but, of course, it also reduces productivity so the total
amount of fume produced per amount of product fabricated is not
decreased. Ensuring various conditions are about right, i.e. arc length,
shielding gas flow, arc current/voltage control (including correct selection of
a.c., d.c. and polarity), is important in restraining fume and noxious gas
production but does not hold the key to really low levels. The gas companies
have for many years believed that the composition of the shielding gas might
hold the key and have devoted considerable amounts of R&D effort to this
problem over a long period of time.
Unfortunately, this is by no means an easy problem: it seems that gas
mixtures that reduce fume increase and vice versa. Ozone and UV
production is reduced by the use of or helium-based mixtures.
However, fumes are in general increased in or helium-mixture welding.
As ozone, iron and oxide particles, and intense UV light radiation are
all potential hazards, the best that can be done today is a compromise.
A discovery that may eventually lead to a solution is the AGA companys
research showing that NO in the welding mix restrained the production of
ozone without appreciably raising levels of The potential toxicity of a
welding mix with NO, albeit at a fraction of a per cent concentration, has
stopped this discovery being generally adopted. More R&D directed
towards a more viable solution is in progress at several gas companies.
Arc/plasma coating. Just as flame coatings can be applied by simply
welding additional metal onto the substrate, so arc welding can be used to
carry out hard-facing of softer alloys. In a similar way, just as flames can be
used to apply coatings formed from metal powder, so controlled arcs can be
used to coat substrates using powdered materials. In fact, the variety of
powders that can be applied with arc or plasma-torch processes is much
larger than can be used with the comparable flame processes. Ceramic
powders and high-melting compounds, such as tungsten carbide, are
possible and the process can be applied not just to metal but also to ceramic
or glass substrates.
Related to arc/plasma coating are processes that aim to build up bulk
articles by thickly coating a mandrel and then separating the final
component. In these processes, of which the Osprey process of Osprey
Metals is one example, a cup-like mandrel or die is sprayed with metal
droplets of size as small as a few microns by an argon or nitrogen spray-gun
fed with liquid metal. The mandrel or die is often rotated to ensure even
deposition. After the die is f ul l , the article is separated from the die, further
heated if necessary and given a final compaction with a hydraulic press. This
process is aimed at achieving high strength in difficult-to-work aero-engine
components. It can produce a metal with a completely even microstructure;
Figure 3.9 Megawatt argon plasma torch and tundish assembly.
no crystal growth patterns relating to casting or boundaries relating to
powder metallurgical origin are seen. Components up to a few kilograms
have been made in this way in Ni/Co/Cr alloys intended for 800C service.
Plasma-torch heating and melting. Megawatt plasmas are used in heating
metal in the steel foundry industry. Steel for casting is prepared in a ladle
(section 3.1.4) and, when its composition is satisfactory, often poured
directly into moulds. However, for continuous casting of strip and bar, a
holding vessel above the water-cooled continuous casting mould (tundish)
is employed. This vessel can be advantageously heated with an argon plasma
torch (Figure 3.9).
A typical apparatus has a lidded refractory vessel with a plasma torch
which can be inserted through a relatively small aperture in the top. The
torch comprises a coaxial copper probe (with water cooling in the walls) with
electrical connections to supply a relatively low voltage (100200 V),
high-current d.c. arc and high melting point tungsten tips. The tips are
treated with thoria to enhance electron emission and assist with arc striking.
High-pressure argon at 10 per hour or more is used to blow the arc 30 cm
or so out of the torch tip, the intense heat melting the scrap charge swiftly
with the mi ni mum of added impurities and exhaust stack emissions.
The big advantages of the plasma torch over other heating methods are
probably convenience (with low capital cost) and controllability. Molten
steel can easily be regulated in temperatures to 5C or so and the small
lightweight torch manipulated through its hole in the top dome of the
tundish (steel-casting melting vessel). When steel is first poured into the
tundish some heat is added to compensate for that lost to the cool
refractories. As casting continues, the level in the tundish goes down and the
charge begins to cool, needing further heating. The addition of a new ladle
of molten steel will require further adjustments in temperature. By
arranging the steel-making ladle to deliver a melt that is always a little cooler
than is required by the chilled moulds, simply adjusting torch power can
regulate precisely the casting temperature. Standard gas burner heating is
less controllable and can cause unwanted carbon, nitrogen or oxygen
additions to the steel.
There are also side-benefits from plasma-torch heating that derive from
the fact that plasma torch systems can be sealed, or nearly sealed, from air
ingress. Oxidation and nitridisation of the melt can be minimised as can
pollutant gas and particulate emission.
Plasmas are also used in the coagulation/melting of metal dust in smelter
and electric arc furnace exhaust stacks. Here fine powdered dusts,
containing large amounts of volatile metals such as zinc, lead, cadmium and
other potentially toxic metals and their oxides, are collected in the usual bag
filters. Instead of dumping, however, the dust is melted by a plasma arc with
micronised coal and fluxing agents such as calcium salts. The melt produces a
relatively innocuous slag, which is tapped off and dumped, and a stream of
reduced vaporised metals that can be recycled to a large extent after
condensing and chemical extraction. It is worth pointing out that some of
these by-product streams from smelters can be extremely valuable.
Companies such as Britannia Refined Metals near London, UK, for
example, which carry out lead smelting, rely on the fact than many lead ores
produces copious quantities of valuable silver.
In the case of electric arc scrap steelmelting, prodigious quantities of dust
are emitted and these are typically dominated by zinc from galvanised scrap.
The volatile zinc is evaporated from the plasma arc furnace and taken off to a
zinc condenser. Zinc recovery can be further enhanced by passing the hot
gas from the plasma furnace through a flame with excess air, forming yellow
zinc oxide, which can be filtered (the zinc-only filter baghouse is much
smaller than that required for immediate filtration of the whole dust stream
and the zinc oxide is pure enough to be saleable).
Similar techniques are used to melt down spent catalysts from automobile
exhaust systems. With the rise in the use of catalysts to restrict CO and
emission from cars, the use of platinum or other precious metals, coated on
to porous ceramics supports, has grown. Although the amount of platinum
in a exhaust catalyst weighing several kilos is only a few grams, it is still
essential to recover it on economic grounds.
Figure 3.10 A robot manipulates a carbon dioxide laser to cut steel automotive parts using
oxygen assist gas (courtesy of LUMONICS Ltd).
3.2.6 Laser welding and cutting
Cutting metals and other materials, from plastic film to refractory ceramics,
is a well-established application of lasers. In industrial practice, a continous
infrared carbon dioxide laser beam is the workhorse of the industry and this
cuts an almost invisibly small slot in metal at a distance of a millimetre or two
from the end of its nozzle.
The use of lasers for cutting relies in general on an assist gas, typically
oxygen, which reacts with the metal melted by the laser and leaves a clean
cut. Similarly, for materials which cannot or should not be burnt, a
high-pressure jet of nitrogen can be employed. The assist gas is supplied to
the cutting point by directing it into a coaxial nozzle assembly. This ensures
the precise placing of gas pressure above the molten metal pool formed by
the laser.
In operation, the first task is to start the cut, either by starting at the edge
of the metal or by keyholing, i.e. cutting an initial hole straight down. The
keyhole is often ragged so keyholing is performed on a part of the workpiece
that will be scrapped. Once started, very heavy gauge and very tough steel
can be cut, leaving only a tiny kerf of wasted material, often less than
0.5 mm. The finish of the cut edge is the key to the economics of laser
cutting; frequently no further operation is required to smooth the edge,
saving considerable finishing costs.
Carbon dioxide lasers, emitting continuous beams of a few kilowatts in the
10 infrared band, are the most common laser used in metal processing.
Neodymium-YAG (yttrium aluminum garnet) lasers, emitting short
extremely intense pulses of 1 radiation, are favoured only for drilling
small holes in thin material. Copper vapour lasers are now being evaluated in
pilot applications for drilling of small holes.
As noted for flame cutting, oxygen purity is an issue with laser cutting. In
laser cutting, the effects of inert argon present in commercial oxygen at the
0.2% level can be seen. Both cutting speed and cut quality deteriorate as the
argon content rises above this level. Pulsing the oxygen cutting gas pressure
rapidly (up to 500 Hz) has been tried as a means of improving cut quality but it
is not clear that any advantage is gained. Laser cutting is now a relatively
popular process and is growing as its quality advantages become widely
known.
Laser welding is still an unusual process. However, it produces very high
quality welds in conjunction with a pure shielding gas, which have the
advantage of no contamination by filler wire or tungsten. Even very thin sheet
metal can be welded; for example, production of disposable steel cans by
welding is now possible, with a joint width and depth of only a fraction of a
millimetre. Helium (or a mixture that is predominantly He) is the gas
favoured for laser welding as its low optical index minimises beam distortion
due to gas heating near the joint and minimises ionisation of the gas, which can
occur at high powers near the focus of carbon dioxide welding lasers.
3.2.7 Ceramic and metal powder forming technology
Powder metallurgy involves the formation of metal or ceramic components
from a precursor powder by pressing into a mould, followed by heating to
bond the particles together (sintering). The precursor powder may well be
mixed with an organic binder, which is burnt out during processing. Most
metals can be prepared in powdered form by spraying molten metal into a
tower filled with inert cooling gas (section 3.2.10).
The sintering process is as follows. First, the mixture of metal powders,
binder and other active substances, such as ceramics, is prepared and pressed
into an oversize mould. The green metal part is then removed from the
mould to a gas-inerted furnace where it is heated to a little (20 or 30%) below
the meltingpoint of the metals used and their alloys. Interdiffusion and partial
melting along the boundaries between powder grains then occurs and the part
shrinks and consolidates. Porosity can be controlled so that sufficient holes
are left that the part can be impregnated with a lubricant such as PTFE
(polytetrafluoroethylene), oil or graphite (a favourite trick for gear wheels
and bearings).
A binder may not be necessary. Small parts may simply use metal powders,
which are then pressed together very hard indeed using a hydraulic press of a
hundred tonnes or more capacity, even for a part weighing a hundred grams.
Gas inerting of the sintering process is essential because of the large
surface area and the consequently potential disastrous effects of even slight
oxidation. In fact, so serious is the oxidation problem that furnace
atmospheres for sintering are typically reducing in nature, with hydrogen
being used or added to the nitrogen or argon inerting. Powder metallurgy
therefore works primarily with metals that do not embrittle with hydrogen.
Examples of sintering are now legion, but a few are given here to illustrate
how sintering is used in industrial practice. Tungsten metal is worked by
sintering. Powdered tungsten produced by chemical reduction of pure
tungsten ore is very hard and brittle (although ultrapure tungsten is softer
and more workable, it is too expensive and unsuitable for industrial
application generally). Tungsten cannot be melted for casting in any normal
fashion and cannot be cold-worked except in thin wires after extensive prior
hot-working: sintering is the only route to bulk tungsten metal working. The
reduced powder is first pressed into preform tubes. These are fired in air up
to red heat then pure hydrogen and temperatures of up to 2000C or more
are used to sinter the powdered compact together well below tungstens
3140C melting point. The still rather brittle sintered rods can then be rolled,
hammered and eventually drawn into wire by raising their temperature to
1000C or more. Tungsten wire made in this way is used to fabricate
filaments for the familiar household incandescent electric lamp.
High performance non-porous alumina products are now made by
sintering. Spray-dried alumina pelletised powder, containing pure alumina
with binders, is pressed, sometimes using isostatic gas pressure techniques
(section 3.2.9) at room temperature, into compact moulds. The compact is
then fired at low temperature in air to drive off binders and form a strong
opaque white ceramic compact very similar to a conventional alumina
ceramic. Finally, in a pure hydrogen atmosphere at approximately 1300C,
the compact is sintered. It shrinks dramatically, by a factor of 30%, and
becomes translucent, non-porous and very strong. A typical application for
this material, which is strong, heat-resistant and oxidation resistant up to
well over 1000C, is for the plasma discharge tubes in sodium high-pressure
lamps (section 3.8.10).
In each of these cases, pure hydrogen is favoured for sintering because it
does not react with the product, diffuses quickly and can be used to form
dense, non-porous sintered products more easily than other less diffusible
gases.
3.2.8 Production of ferrites and ceramic high critical temperature
superconductors
Ferrites are magnetic minerals containing the ferrite anionic group. They
are unusual amongst strongly magnetic materials in that they are good
electrical insulators. This enables them to be used in high-frequency circuits
with low eddy current losses. Simple ferrites have the formula
where M is a metal ion or mixture of ions typically Mn, Ni or Zn and less
frequently Mg, Co or Cu. Strontium and barium hexaferrites, are
examples of ferrites with a more complex crystal structure.
Ferrite powders are made by fine ball-milling a mixture of oxides then
reacting, calcining, followed by re-milling of the resultant sinter. The
powders can be formed into artefact shapes by green forming. As with
other ceramics, powder is mixed with binders and solvents until it forms a
plastic mass which can be moulded. The moulded green preforms can then
be heated slowly to drive off solvents and burn off organic binders. They can
be fired in oxygen to achieve high densities but are more commonly fired in
hydrogen or inert gases when cracking due to excessive formation of
iron(III) oxide is avoided.
The production of working copper-oxide based ceramic superconductors
was worked out very recently; first results were reported only in 1986.
Superconductors have a number of unique properties, the most important of
which is that they can conduct direct current electricity without any loss, i.e.
they have zero resistance, which is a highly desirable property in electrical
engineering. Most superconductors only achieve this state of electrical
nirvana at temperatures a little above absolute zero. Above this critical
temperature they revert to being ordinary electrical conductors. They
thus need liquid helium as a coolant, which is an expensive technology,
restricting application of superconductors almost entirely to esoteric
applications where huge magnetic fields generated by huge zero resistance
electromagnetic coils are needed.
Ceramic superconductors were valued at first for their high with many
compounds being developed to have a above the 77 K of liquid
nitrogen. At these temperatures refrigeration is simple and inexpensive,
making them very attractive. In addition to high however, many of the
cuprate superconductors also possess a property which may turn out to be
more important than their high they retain their superconducting state
even in ultrahigh magnetic fields.
Both the and the maximum magnetic field of a high material
are very dependent on how the superconductor has been prepared. Typical
superconductors are composed of copper oxide with rare earth metals,
such as YBCO (yttrium/barium/copper oxide) and BISSCO (bismuth/
strontium/copper oxide). A typical preparative method would be to grind
the metal oxides together very finely in a ball mi l l , followed by a reaction
firing or calcining at a high temperature. Other methods start with
precursors which are water-soluble salts, carbonates or acetates, which are
mixed in aqeous solution, evaporated and pyrolised in oxygen before firing.
The product of this firing is partially reacted ceramic, which is then ground
up into a fine powder again and put through a forming process to achieve the
required artefact shape. The forming processes employed vary from packing
powder into a silver tube and detonating explosives all around it, to hot
isostatic pressing (HIP) and the so-called plastic ceramic route. The latter
involves mixing the powders with organic solvent and binder compounds,
followed by plastic forming of the resultant black dough into the required
shape. The organics are removed by a low-temperature pyrolysis step.
Next comes the vital firing process in a precisely controlled oxygen-
containing atmosphere. During this firing, a complex layered crystal
structure of H ceramic minerals forms.
3.2.9 Hot isostatic pressing
Hot isostatic pressing of powdered preforms, whether of metal or ceramic, is
now an established industrial process. Instead of pressing a powered
preform between two mechanical anvils, the preform is placed in a very high
pressure vessel while this is filled with ultrahigh pressure gas (up to a few
thousand bars). The preform is then heated. Hot isostatic pressed materials
often have improved properties; mostly incrementally improved but
occasionally spectacularly better.
Many metal-working processes involve consolidating more or less porous
metal artefacts. Mostly this is done by the application of hundreds of MPa
(few thousand bar) of compression using a hydraulic press with a suitable die
at an elevated temperature. However, it is difficult to make anything but
simple components in this way and the pressure applied is essentially
uniaxial in nature. With uniaxial pressure, any pores that may remain are
disc-shaped and more liable to cause failure in the component. Also, the
control of temperature is difficult unless the die and press jaws are heated
and temperature controlled, which is difficult in a simple press.
Originally developed primarily for the processing of reactor fuel rods
(diffusion bonding of cladding to uranium fuel) and then used extensively
for tungsten carbide, HIP spread in the late 1970s to many other applications
in the area of reducing porosity in complex castings and consolidation of
components produced by powder metallurgy. A typical HIP process
involves the following steps:
encapsulation or coating of precursor object;
placing of object in pressure vessel, evacuate or purge;
filling of pressure vessel to 50 or 100 MPa with pure argon;
raising temperature of object and gas to around 0.7 times its melting point
at constant volume (this step causes a further pressure rise, up to hundreds
of MPa).
The essential idea of HIP is to apply heat and an isotropic compression to
a partially porous component. Ultrahigh pressure pure argon is a convenient
means of delivering that heat and pressure. At these temperatures particles
of metal or ceramic bonded only at their extremities become subject to
diffusion of atoms through the bulk of the object, especially along grain
boundaries and along the porosity surface, the process of sintering is taking
place. In an ideal case, pores shrink because the temperature allows atoms
to move and surface tension tries to shrink the pores. However, sintering at
low pressure leaves pores, often spheroidal gaps which have filled with gas
and have an internal pressure, preventing their complete disappearance.
Components produced by compaction and sintering of metal or ceramic
powders will always tend to contain pores. Similarly, most metal castings,
because of crystal grain growth and shrinkage problems, are slightly porous.
High levels of trapped gases such as hydrogen or carbon monoxide, often
present in casting metals, further exacerbate these tendencies. Only by the
thorough removal of this residual porosity can full strength be attained.
The application of pressure in addition to sintering temperatures enables
the complete removal of porosity. The pressure applied will completely
swamp any residual gas pressure and enable the gas from remaining pores to
be dissolved.
The HIP process is not a metallurgical panacea, however, as it has a
number of problems:
the vessel needed is expensive and potentially dangerous;
the compressor needed, generally a diaphragm-isolated reciprocating

type, is expensive to buy and run;
the cycle time is necessarily fairly long (usually a few hours but the vessel
size, a metre or so diameter, limits the batch size);
surface-connected porosity is not compacted but is simply consolidated as

if it were part of the component design.
The latter point was a serious objection to the application of HIP to castings
with surface defects and powder components until suitable coatings and
encapsulation were developed. Thin sheet metal, usually steel, is often used
for powder metallurgical components. Ideally the sheet covering is pressed
into shape, the component sealed inside under vacuum and removed after
HIP by acid dissolution or machining. A potentially simpler alternative is to
coat with a slurry of ground glass. The glass type is chosen to soften at HIP
temperature; it must be fused before HIP pressure is applied. There is much
research into other surface-connected pore treatments, such as metal
deposition by plating or sputtering.
The argon gas needed for HIP is usually recycled on frequently used
systems but must be carefully purified before reuse; a typical specification
might be 99.99% purity. This is because even small partial pressures of
impurities such as oxygen can lead to serious oxidation in the components or
furnace. The diaphragm compressor is preferred because there is then no
question of the compressor adding impurities to the gas stream. The argon
gas should ideally be checked for common contaminants, such as
and with trace gas analysers.
Argon is used above nitrogen or other gases because it does not dissolve to
any appreciable extent in most materials and hence there is little danger of
argon bubbles forming on depressurisation. Helium can also be used,
although at some penalty in cost. At the conditions used argon gas has a
density comparable with that of water. At these unusual densities, the gas is
an excellent heat transfer medium and HIP processes normally rely mostly
on this convective heat transfer rather than radiation.
HIP of materials which dissociate to some extent in an inert atmosphere at
elevated temperatures can be aided by mixing some of the dissociation
product with the argon. For example, addition of small amounts of oxygen
or nitrogen can aid in the production of oxide or nitride ceramics.
A typical HIP installation comprises a pressure vessel with heaters and a
circulator with safety systems for leaks and pressure vessel cracks. A
diaphragm-isolated reciprocating compressor is used to pump the argon,
which is often repurified and recycled after use. Facilities for the preparation
of samples, such as an atmospheric sintering furnace, powder compaction
and encapsulation, are also required.
3.2.10 Metal powders by spraying
A lot of metal powder is used as a powder, for example in metallic paints, as
a filler for plastics, for catalysts and so forth. However, a major application
for metal powders is for applications where bulk metal is the ultimate object,
such as the preparation of sintered metal objects. For sintering, it is a big
advantage to have a free-flowing and easily compacted metal powder, and
this can be achieved by using metal powder comprised of more or less
spherical particles.
There is a long history of manufacture of spherical metal particles of larger
size for use in shot for sporting ammunition. Lead or lead/arsenic alloys were
poured or sprayed into the air inside a shot tower a few tens of metres
above a quenching pond. Surface tension forces ensures that approximately
spherical particles of a few millimetres in diameter are formed, which cool
sufficiently in the air of the tower that they will not be distorted when they hit
the cooling water. The formation of spheroidal particles of metal for powder
metallurgy follows similar lines but uses an inert gas for the cooling medium
and no pond.
A typical installation pumps a liquid metal stream into a jet of nitrogen or
argon at up to 40 bar breaking up the metal stream into small droplets. The
atomised metal, from a few microns up to 100 m particles, can be allowed
to cool in a free fall through stationary inert gas or in recirculating gas blast,
from which the particles are separated from the gas. The result is small
spheroidal particles quite unlike the angular particles produced by simple
crushing and aqueous atomisation followed by evaporation or oxide
reduction. A large atomisation unit is a big user of inert gas; figures up to
several hundred cubic metres per tonne of metal powder are quoted for
some installations.
Atomised metal powders in finer grades are often passivated. Unpassi-
vated fine metal powders, even of such relatively stable metals as nickel,
ignite very easily in air and may even be pyrophoric, i.e. they spontaneously
ignite on being poured through air, such as iron filings emitted from a
grinding wheel working on steel ignite. Gas mixtures containing small
amounts of oxygen are often used to passivate pyrophoric powders. Such
fine powders are most often used as catalysts; the hydrogenation of
vegetable oils, for example, is mostly based on finely divided nickel
powders. One way of avoiding the handling dangers that this entails is to
slightly oxidise the surface of the powder, by drying it in nitrogen or similar
and then blowing nitrogen or argon with 1 or 2% oxygen added through it so
that a thin oxide film can form with no danger of combustion. The resultant
powder is thus rendered non-pyrophoric.
Finally, it is not always necessary to isolate the powder production from
solid object production. In some processes, typified by the Osprey process,
an aerosolised liquid metal stream made with argon or nitrogen is formed
directly into a solid object of preform by spraying the particles, while they
are still hot, onto a rotating or otherwise moving mandrel.
3.3 Industrial gases and chemical manufacture
3.3.1 Oxygen for catalytic converters
The use of oxygen in many petroleum processes is now routine. Most large
refineries have a large on-site air-separation unit, producing both nitrogen
and oxygen. Within the many potential applications for air gases around the
refinery, selection is based on economic criteria.
One often selected is the addition of oxygen to catalytic cracker
(cat-cracker) units. These units are designed to convert much of the heavy
hydrocarbons in crude oil into lighter fractions suitable for motor fuel.
Cat-cracker units vary in design but one type functions as follows.
Zeolite-type silicate catalyst, ground to 100 m or less, is mixed with heavy
crude oil and the resultant slurry heated to several hundred C. Lighter
fractions are produced, as required for automotive petrol and for liquid
petroleum gas (LPG), along with very heavy hydrocarbons stuck firmly to
the particles of zeolite. The zeolite is then regenerated by removing as much
volatile material as possible by steam stripping, followed by combustion in a
pressurised air stream in the catalyst regenerator unit. Waste heat from this
process is used to preheat the new slurry whilst the burnt catalyst is
regenerated and can be mixed with more crude and reused.
If cat-cracker units are already working to their design capacity, a
relatively simple way to increase their capacity is to debottleneck them by
boosting their throughput with oxygen injection. The oxygen enrichment of
the burn-off air reduces the amount of coke deposited on the catalyst during
the cracking operation. The large volumes, up to hundreds of tonnes per
day, used in this application mean that it is one of the larger uses in the
chemical industry.
3.3.2 Gasification of coal
The most plentiful fossil fuels on the earth are the various sorts of coal.
However, coals are inconvenient sources of energy because they:
are buried deep underground;
produce copious waste (noxious sulphur dioxide, unpleasant ash) when

combusted;
are highly variable in quality;
are solids and therefore are expensive and difficult to handle both in
transport and inside engines and boilers.
The efficient production of a gaseous fuel from coal is, therefore, a
desirable objective, eliminating, at least for the user, all the problems of
using coal and leaving only the problem of mining it. This fact was also plain
in the nineteenth century when gasworks were installed at most industrial-
ised towns of any size. However, the gas they generated (a mixture of easily
ignitable hydrogen and toxic CO) is no longer acceptable for domestic
consumption.
49
The production of high-grade methane fuel gas from coal is not an obvious
use for oxygen. The basic reaction desired is, after all, the gasification of coal
with water
However, most processes for substitute natural gas (SNG) from coal do
involve oxygen, with most processes, perhaps surprisingly, needing more
than a tonne of oxygen for every tonne of methane produced. The apparent
anomaly is because this direct process is not in general possible, the above
reaction is endothermic and heat must be obtained from further coal
burning, so the overall reaction scheme is approximately
The additional combustion is best carried out with oxygen because it does
49
That said, early gas appliances, including Bunsen burners and the like in the laboratory, were
still running on coal gas in the late 1960s and safer methane would also have caused problems
in many situations. When methane first replaced coal gas in UK, there were a number of
incidents where burners failed to ignite, owing to the much higher ignition energy of methane,
causing potentially dangerous accumulations of fuel gas.
not introduce further unwanted gas into the reactant stream. Alterna-
tively, the combustion process can be separate from the gasification step, the
heat being transferred by conduction or transfer of hot uncombusted coal,
for example, so that nitrogen from the air is not present in the gasification
step. However, a typical process does use oxygen.
A typical gasifier process, in outline, is as follows. Pellets of coal are added
steadily to the top of a reactor running at tens of bar pressure, the pellets
reacting and sinking down whilst a counter-current of hot gas flows upwards.
Reactions are zoned, with combustion at the bottom heating the coal to as
much as 2000C in a stream of steam and oxygen. In the cooler middle region
CO is formed by the reaction
Also in this region some methane is formed by the endothermic gasification
reaction. Finally, in the top zone more methane is formed by hydrogenation
of tar in the coal whilst the new pellets of coal are preheated. The overall
reaction in a typical gasifier falls short of the ideal in that the calorific value of
the outlet stream is rarely over 50% methane. Some additional upgrading
process, e.g. a catalytic reactor, is required to shift CO and into more
methane.
Considerable variations in SNG processes are possible, with temperatures
from 1200 to 2000C in the combustion zone, for example, and the coal used
varies from soft, highly impure lignites to the almost pure carbon of
anthracitic coals. This latter variability is a problem, as is the current low
price of natural methane. However, if SNG is required as a real alternative
fuel, then very large oxygen plants may be needed whatever the type of
process and coal. 5000 tonne per day plants are indicated for commercial
scale SNG.
3.3.3 Nitrogen and injection in oil wells
The gusher is still, in the minds of most people, the stereotype of an oil
well. A deep hole is bored in the right part of Texas or Arabia and when the
drill finally penetrates the oil-bearing rock oil magically squirts out of the
hole, soaking the rig operators in their black gold.
In the early days of petroleum the large hydraulic and gas pressures
present in many oil wells were commonly relied on for initial extraction. As a
well began to run out of its own natural pressure, pumps, often the classic
nodding donkey pumps, were applied. All oil used to be produced by
gushers and pumping. Today, however, when it is necessary to ensure a
much higher percentage of the oil in a reservoir is recovered, injection is
common.
Injection into an oil well, by pumping into neighbouring boreholes, was
first carried out mostly with water, often treating the water with chemicals to
avoid plugging of the well. Gas injection is also now common. Often
hydrocarbon gases are injected. However, it is cheaper to inject or
nitrogen into a formation. Nitrogen has the advantage that it is only slightly
soluble in oil and in rock, and so smaller volumes can be injected than would
be the case with and hydrocarbons. A non-oxidising gas is needed for
injection; air injection, at the high pressures and temperatures involved,
would lead to partial combustion down the drill-hole. The latter has been
tried for in situ gasification of petroleum, tar and coal but is experimental
compared to the straightforward gas injection to enhance production that is
now standard in the industry. PSA equipment is often used for making the
large quantities of nitrogen needed. The low cost of gas produced on-site by
PSA is the key, whilst the typically low purity of the PSA plant is not
important.
3.3.4 Oxygen for oxidation of chemical intermediates
The preparation of many oxygenated solvent and intermediate products
relies today on the catalytic reaction of oxygen with unsaturated hydro-
carbons in the presence of a transition metal or precious metal catalyst at a
few hundred C. Oxidation occurs via an intermediate oxidant whose state is
restored by reaction with gaseous oxygen. Acrylic acid, ethylene oxide,
formaldehyde, phthalic anhydride, vinyl acetate and vinyl chloride are all
susceptible to manufacture by catalytic partial oxidation.
The use of oxygen, as opposed to air, for these oxidation reactions
depends primarily on economic factors such as size and therefore capital size
of plant. With the more intensive reactions achieved with oxygen, smaller
plant is needed with a lower capital cost. Where more production is needed
from an existing plant, the use of oxygen enrichment is often a possibility,
with minimal capital investment. Provided the cost of the oxygen is
exceeded by the value of the extra product less its other marginal costs,
therefore, the use of oxygen is justified.
Other economic factors are the cost of heating or cooling the unwanted
nitrogen stream, even if this can partly be recovered by heat exchangers. If a
process needs compressed air at a few bar, then there is again wasted energy
involved in the compression of the unwanted nitrogen. Indeed, the
compression of air to a few bar is all that is required for the operation of an
air-separation plant producing pure oxygen. Similarly, the running costs of
scrubbing plant to extract the product and recycle the gases are reduced with
oxygen feed.
There may be strong technical, rather than simple commercial, factors
favouring the use of oxygen, however. Classic cases of this kind occur when a
process does not carry a reaction close to completion. In order to use the
unreacted gases effectively, they must be recycled back to the input.
However, if air has been used, nitrogen and argon will build up in the system
and will need to be vented or extracted. This would cost more valuable
energy or materials through the need for venting or separation of the
nitrogen, as otherwise the reactants would be diluted, with a resultant loss in
plant productivity. The BOC company has patented several PETROX
processes which combine a pressure-swing absorption unit with oxygen
enrichment to carry out reactions such as butane to maleic anhydride
(oxidation) or propane to acrylonitrile (propane dehydrogenation then
ammoxidation using and ). These constitute an efficient process by
recycling the unreacted alkane back as input feedstock to obtain high yields.
Similarly, nitrogen oxide production occurs in all high-temperature plant
using air feed; this can be avoided by using oxygen.
Some processes are unsuitable for use with pure oxygen because they
proceed too vigorously, whilst for others the plant size is already small. The
wider explosive limits for gas/oxygen mixtures also restrict the applications
in which oxygen can replace air in gas-phase reactions. It may even be
necessary to dilute air with nitrogen to reduce the oxygen content of the
reactant stream; another potential application for air-separation gases.
Reactions in aqueous media do not generally, of course, have this problem.
An example of a process which is simpler and cheaper with pure oxygen is
the HoechstWacker ethylene to acetaldehyde conversion. This uses a
mixture of aqueous palladium and copper chlorides to effect the direct
oxidation
The conversion is very efficient at a few bars pressure and 100C. The
simplest apparatus involves a reactor followed by a water scrubber. The
liquid from the scrubber is distilled for the product. In a similar way, it is
economic to use oxygen for the production of ethylene dichloride, ethylene
and propylene oxides and vinyl acetate.
There are further economic factors involved in the full use of all the
products of air separation. If a chemical complex has many uses for nitrogen,
for inert blanketing, flammable feedstock and product tanks, for example,
then the use of oxygen in a process may be favoured.
3.3.5 Oxygen-enhanced sulphur recovery
The Claus process is one example of several patented processes for reducing
the sulphur content of petrochemical streams. It uses oxygen injection to
both remove sulphur from the process stream and to recover tolerably pure
sulphur. The oxygen oxidises the hydrogen sulphide in a controlled manner
so that elemental sulphur is produced. If is burned in the normal way
with oxygen the result is sulphur dioxide. Early Claus plants simply took one
third of the input stream and combusted it completely then mixed it with the
original stream where the reaction
produced elemental sulphur. Claus recovery plants typically work on gases
containing , , . This is the sort of mixture which results from
treatment of fuels such as natural gas for acidic gas impurities. A liquid
absorber takes up the and , leaving, in the case of natural gas, pure
methane. The absorber liquid is then transferred to another vessel where it is
heated and depressurised to desorb the gases after which it can be recycled.
The Claus plant passes this mixture of desorbed gases through a catalytic
reactor. Recovery factors for sulphur content vary up to about 97%.
Arguments for the use of oxygen rather than air for the oxidation revolve
mainly around the larger size and therefore higher capital cost of air-based
plant versus the cost of oxygen.
3.3.6 Ammonia and fertilisers
The production of ammonia often proceeds industrially by the most direct
route possible, i.e. the Haber process. Nitrogen from large on-site
air-distillation plants is added to hydrogen from large steam reformers. This
reaction is exothermic but requires extraordinary pressures of approxi-
mately 200 bar. After multiple stages of compression via enormous banks of
axial and centrifugal compressors, the gas mixture is reacted at 200 bar or
more. Argon is one of the by-products of the ammonia business. This arises
as follows. Nitrogen must be produced in large quantities, along with
hydrogen, in the ratio 1:3, according to stoichiometry of the reaction
The hydrogen is generally produced by reacting steam and methane in the
steam reformer reaction
However, this reaction needs high temperatures and is endothermic. The
heat and nitrogen needed can be supplied by adding air to the reaction
stream, oxidising further methane to CO exothermically. This is ingenious
as it is not necessary to separate the nitrogen from the air
A catalyst, e.g. iron oxide at 500C, will convert the CO to if more
steam is added, leaving only which is easily scrubbed out with an
absorber tower, and . The stream is supplemented as appropriate
from the steam reformer reaction to yield a 1:3 mixture.
There is a snag, however, to the cunning use of air in this way. The
resultant stream contains a fraction of one per cent of argon. This
would not perhaps be too much of problem were it not for the fact that the
mixture is not reacted in one go into ammonia but is recycled several
times, depending on process conditions and catalyst, each time scrubbing
out more ammonia but leaving increasing amounts of argon. Eventually, the
argon builds up to tens of percent and must be wasted to wash out the
inert argon. Alternatively, if the process is operated to maintain the argon at
a compromise level in the , then Ar can be commercially extracted in a
cryogenic plant. Too high a level of Ar will inhibit the ammonia synthesis,
too low a level will make the Ar extraction too expensive.
Most ammonia is made into fertiliser. After reaction with nitric or
sulphuric acid, and the admixture of phosphates or other components, the
aqueous fertiliser is sprayed into a prill tower. Here an updraught of warm
air carries the droplets upwards. They rapidly dry not into crystals but into
small prills i.e. free-flowing rounded granules a few millimetres across.
They are collected from the top of the tower and then bagged for customers.
Ammonium nitrate is also important in the manufacture of nitrous oxide
gas.
3.3.7 Inerting chemicals
Inerting pipework. As noted above (section 2.5.10), before a pipeline is
filled with a flammable or reactive gas, it must be purged. To prevent fire and
explosions in large pipes carrying inflammable gas or vapour, pipes are
purged with several volumes of nitrogen before and after use, ideally with
pressure cycling. Reactive gases and pipelines needing high purity may
require much more purging than this using all the purging tricks detailed
above.
The purging is often carried out far beyond safety requirements to ensure
quality. Silane in electronics applications, for example, is useless if even a
few ppm of oxygen creeps into it; the reacts to produce paniculate silica
(i.e. a rather fine sand), which will deposit on the microcircuit being
fabricated and ruin it. Similarly, large stores of resins and paints are often
much more carefully inerted and controlled than fire safety would dictate
because even small amounts of oxygen from the air lead to oxidative
polymerisation, i.e. the resin thickens and skins, leading to problems later.
Inert blanketing of tanks. The ullage space above tanks containing liquids
which are flammable or reactive to oxygen or other components in air, such
as moisture or carbon dioxide, is best maintained with an inert gas. As
noted, many paints will react to set prematurely if oxygen reaches them,
whilst dessicating agents and cyanoacrylate glues must be kept in an
anhydrous condition to be saleable. Still other chemicals can usefully be
inerted because of materials considerations. When maintained oxygen-free
and anhydrous, many chemicals, even acids, are much less reactive to-
wards inexpensive constructional materials such as mild steel.
With liquids forming a flammable vapour, the breathing of a tank can
result in vapour being lost via vents during filling or temperature increases,
with the risk of a flame at the vent. Worse still, on emptying or cooling, the
tank can inhale air to form an explosive mixture in a confined space, which
is clearly extremely dangerous.
The simplest way of ensuring blanketing is to simply connect a steady
stream of inert gas (almost invariably nitrogen) to the top of the tank. This
stream is then vented continuously. Provided the product withdrawal rate
from the tank does not exceed the inert gas input rate, the tank will remain
under ful l y inert conditions. A more sophisticated variant is to add another
valve which switches to a higher flow rate when the tank is being drained
and otherwise uses a low inerting flow. This method, although simple,
leads to higher gas usage and loses a small amount of product vapour con-
tinuously (see Figure 3.11a).
Alternatively, sensitive gas regulators can be employed so that when
liquid is pumped in or out of the tank a slight positive pressure of gas, a few
millibars on a water gauge, can be maintained above it by venting or
adding gas. The slight positive pressure ensures that any small leaks will be
outwards in direction while putting minimum strain on the tank. Product
vapour is only lost during product filling. Even this loss can be recovered to
some extent by the use of carbon absorbers. These absorb product vapour
when the tank is venting and can be stripped at least partially of that
vapour by backflushing into the tank. A more sophisticated and more ef-
fective vapour recovery system uses cryogenically cooled absorbers to
further increase efficiency. At least for very large tanks, in this kind of
vapour recovery system inert gas is only used during product withdrawal.
Inert gas can be shared between tanks on large-tank farms, provided slight
admixtures of vapours can be accepted, further reducing the quantity of
inert gas needed.
A scheme for implementing tank inerting in this way is shown in Figure
3.11b. Two ordinary regulators reduce the house nitrogen line pressure
down to a few hundred millibars or less when a sensitive and accurate low-
range final regulator is used. Check valves are included to avoid the possi-
bility of solvent entering the nitrogen system and pressure-relief valves are
supplemented by a vacuum-relief valve to prevent the tank collapsing.
Flame arresters on the vent pipes are good practice that avoids a flame
spreading to the ullage.
During commissioning, the nitrogen regulator is set artificially high so
that the large flow rate can purge out the ullage. In operation, the nitrogen
regulator is set a few millibars lower than the two relief valves, giving the
system some hysteresis, which will avoid unnecessary hunting of the system
control and excess venting losses.
Figure 3.11 Balanced pressure tank inerting system: (a) tank blanketing with continuous
nitrogen purge; (b) tank blanketing system with semi-closed circuit.
Nitrogen foam inerting. Nitrogen can be used to generate inert foams,
which are especially good fire suppressants and extinguishants. Such foam
has often been employed when modifying or demolishing chemical plant
that has, or used to have, flammable materials inside it. Specialist liquid
nitrogen trucks are provided with a propane-fired vaporiser along with a
pump for water and a tank of medium- or high-expansion foam concentrate.
A foam gun combines these ingredients and can be used to make up to
of foam per minute.
The typical job for such a foam generator is during demolition or repair of
tank farms and other chemical plant. Tanks that contain small amounts of
residual flammable liquid will be more or less full of flammable vapour,
which, if air is allowed to enter, constitutes a severe explosion risk. During
demolition, tanks can be filled with foam and then cut up to leave a
wide-open top. An oxy/acetylene torch can be used and foam leaking out of
the cut will not extinguish the cutting flame. Once the top is off the tank it
can be left open and demolition completed after the foam has drained away
and freely circulating air has removed any vapour. For repairs, the tank is
drained (although sometimes this is not necessary) and foam fed in. The hole
or crack can then be repaired by electric or oxy/acetylene welding in the
usual way and the foam drained off. After that the tank is washed and
refilled ready for service.
There may be problems with some solvents which react with and disrupt
the foam stability but often choosing a different detergent or soap can
handle this. Most solvents and oils are less dense than water and it is
therefore important to drain off excess liquid from the bottom.
3.3.8 Miscellaneous chemical applications
Pressure transfer and pneumatic lift. Highly corrosive and solvent liquids
and, in the semiconductor industry, ultrapure liquids can be difficult to
pump. The rotating or reciprocating seals in many pumps add particulates to
ultrapure liquids and even the minutest leaks of corrosive materials past
seals can cause disaster. Modern diaphragm pumps, often fabricated almost
entirely from Teflon and related perfluorinated polymers, have overcome
many of these problems but are expensive, give a pulsating flow and often
require regular maintenance, which can be awkward in confined locations.
Using gravity is often possible but placing tanks high enough to give suitable
dispense pressures is often difficult and undesirable for large quantities of
hazardous liquids.
Two solutions which take the radical approach of removing pump
problems by removing the pump, are pressure transfer and pneumatic lift.
Pressure transfer refers to the method of dispensing liquid from a tank by
applying gas pressure to the ullage space. Pneumatic lift is where gas is
bubbled into the bottom of a liquid pipe that is taken well below supply and
outlet level. The gas bubble/liquid mixture column has in effect a lower
density than a liquid column and thus the liquid can be carried up to
considerable heights.
Pressure transfer requires a vessel capable of handling a reasonable
pressure (often dubbed a pressure egg). Incoming liquid is delivered into
the pressure egg by first venting the vessel then admitting it. Once the
correct level is reached, the vent and inlet valve are closed and nitrogen
pressure applied. Whenever the dispense valve is opened, the product is
dispensed under pressure. With good pressure control and electronic timing
of the valves, precise quantities can be dispensed into a downstream process.
The level in the pressure egg can be measured in the usual way, e.g. by a
Figure 3.12 Pressure transfer system.
differential pressure indicator or transmitter, or simple on/off level indi-
cators can be used.
Some applications lend themselves to pressure transfer. Highly viscous
materials, such as glues and molten sulphur, cause problems in pumps,
particularly as it may not be possible to heat the pump to the same
temperature as the tanks and pipeline. In these cases a pressure transfer
system is especially appropriate.
Pneumatic lift systems are usually designed to transfer liquids from one
atmospheric pressure tank to another, the latter being higher than the
supply tank. They rely on a pipework system which goes considerably below
the supply tank. Gas is injected into the pipework below the supply tank at a
position (1 m or so into the up-going pipe) where bubbles are sure to rise
towards the receiving tank. The gas and liquid in the up-going pipe then
constitute a column of material of lower density than that in the down-going
pipe. The pipe is chosen to be of sufficiently small bore that the gas cannot
simply bubble up the column, but the liquid and gas form a genuine mixture
(although there will always be some slippage, i.e. gas flowing faster than
liquid). A hydrostatic equilibrium is soon established with the lower density
column displaced upwards so far that liquid overflows out of the up-going
pipe and is delivered to the receiving tank.
The gas/liquid mixture in the up-going pipe, at reasonably high rates of gas
injection, shows the bubble amalgamation effect seen in many bubbling
apparatus (section 3.6.1) and the flow pattern soon settles down to a slug
flow. In this, slugs of water, interspersed with large bubbles that fill all the
pipe except for a thin film of liquid, flow upwards. At higher gas flows the
slugs are smaller and more irregular, and the term churn flow is sometimes
used. Unstable hunting of the flow rate is often seen, related to
acceleration of liquid in the pipes. This can be controlled to a large extent by
incorporating a check valve somewhere in the pipework, thus preventing
backflow.
The depth at which the gas must be injected can be calculated as follows.
At hydrostatic equilibrium the pressure supporting the supply and receiver
vessel fluid columns of density and , respectively, must be equal, so
or
where is the minimum submergence (distance of gas injection point
below supply tank), is the difference in tank heights, is the density
ratio '/ and g is acceleration due to gravity. Clearly, due to frictional losses
and other non-ideal effects, D must be made substantially more than It
can be seen that if is only, say, 0.8, only a little gas is injected. In this case a
pit or borehole may have to be dug to allow D to be large enough ( =
). On the other hand, if it is permissible to use a large amount of gas
injection, then = 0.4 might be usable, giving = 0.7
The mean density in the slug flowing pipe is the key factor in the
calculation of whether a particular configuration is workable. This can be
estimated for water and other common fluids by scaling from empirical
measurements in the literature. For unusual fluids (the effect of dissolved
surfactants or viscosity modifiers are of course important) there is no
substitute for a small trial. However, it will often be found that a mean
density of 0.5 or so can be achieved (assuming the gas density is negligible
compared to that of the liquid). A warning should be sounded, however,
that achieved density varies with flow rate of liquid as well as that of gas and,
perhaps surprisingly, at low liquid rates, because of the change of flow
conditions in the upgoing pipe, density may not fall with increasing gas rate.
Older standard textbooks (see, for example, Coulson and Richardson,
1965, 1970) often give the full theory of pneumatic lift pumps. It is sufficient
to say that they can be made very efficient (50% in terms of usage of the
energy of the compressed gas) provided a deeply submerged pipework
system can be managed and very high flows are not required.
Both pressure transfer and pneumatic lift can be carried out with many
substances in air. However, the very reactivity of many chemicals for which
the methods are useful and the maintenance of ultracleanliness in semicon-
ductor work, mean that using inert nitrogen gas is often the method of
choice. Nitrogen avoids problems with oxidation of product, fire and
explosive vapour hazards, and is not contaminated with other potential
Figure 3.13 Pneumatic lift system.
reactants such as carbon dioxide and water vapour. Triboelectric charging
phenomena often occur, causing large potential differences between
insulated vessels containing non-conductive liquids. Sparks produced by
these voltages can cause ignition in an unexpected way, so nitrogen inerting
can be particularly valuable.
3.4 Gases in the food industry
3.4.1 Freezing and chilling
The freezing and chilling of food using liquid nitrogen is now very big
business. In many countries it is the biggest user of tonnage liquid (as
opposed to gaseous) nitrogen. The total world market for frozen food is now
of the order of $100bn. At first, in the 1880s, crude techniques such as brine
immersion were used. Business pioneers such as Clarence Birdseye
developed economic freezing systems in the 1930s using cold-plate freezing.
The much more rapid mechanical air-blast and the even faster cryogenic
techniques followed soon afterwards, along with a shift towards higher
quality IQF (individual quick frozen) products. IQF foods are preferred by
consumers as the quick freezing gives a tastier, better textured product and
the separate pieces of produce means just the right amount of product can be
defrosted for use.
To preserve it, food must be stored well below the freezing point of water,
around 18C giving a reasonable lifetime. Simply placing food in a cold
store at 18C and hoping for the best, however, is not in general sufficient.
Under these circumstances it could take a day or more even for a small (e.g.
a few pounds) package of typical food to freeze completely and cool to near
18C with natural convection alone. This is so slow that some changes will
occur in the food. There would be time for some bacterial growth or other
spoiling effects and therefore food that is frozen slowly does not taste fresh
when thawed. Slow freezing may allow ice crystals to form inside the food,
ruining its texture when thawed. Thus there arises a need for a specific food
freezing process prior to storage or distribution as frozen produce.
The freezing process. Normally freezing of foods is done by use of
large-scale, but otherwise rather conventional, refrigeration plant. A
working fluid such as a CFC (or HCFC for newer plant) or anhydrous
ammonia is compressed and then cooled by an air or water-cooled heat
exchanger back to room temperature, at which point it liquefies. The liquid
refrigerant is then allowed to expand at a nozzle then cools, absorbing heat
from the cooling plates inside the plant and boiling-off. The gaseous
refrigerant is then recycled back to the compressor.
In order to effect rapid freezing of foods, the conventional plant is run at a
very much lower temperature (30 or 40C, or lower) than a normal cold
store and large fans are employed to blast the air rapidly past the food being
frozen. This is the more or less standard air-blast freezing machinery used
in many freezing plants.
The time taken to freeze food with this sort of modern freezing machinery
is, however, still quite lengthy. A 10 kg package, for example, might take 4
to 8 h to freeze and cool down acceptably. Some acceleration of this is
possible by processing smaller packages and by using a higher speed air
blast; air speeds of up to 10 m s
1
are used. With mechanical freezing, even
with these powerful icy blasts, there is still a distinct pause phase in which ice
crystals form. The initial cooling to a degree or two below 0C is rapid, as is
the final cool-down to 40C, but there is a long plateau. It is during this
plateau phase that growing ice crystals can cause damage and evaporating
water from the product surface can leave it dehydrated. The freezing plateau
can also lead to drip loss. This is an effect caused by osmosis from the plant
or animal cells into the intercellular space. As the ice crystals form in the
intercellular space, the liquid remaining there becomes concentrated and an
osmotic pressure differential is built up which causes the cells to leak water.
Subsequently, both cells and space are frozen and further osmosis stopped.
On thawing, however, there is more water (up to 5% more) in the
intercellular space, some of which simply runs out of the product as drip loss.
Drip loss is the percentage of the produce lost in watery fluid as it is thawed
before consumption.
Liquid nitrogen (LN) offers the potential for very rapid freezing. This is
important because it means that many foodstuffs which contain a lot of free
water can be frozen. If frozen slowly, the water will crystallise into long
spicules which break the cell walls within the foodstuff. When this frozen
produce is then warmed up to room temperature it will be mushy because of
the destruction of the cellular structure which normally gives it texture.
Frozen produce with this kind of problem will also suffer excessive drip loss.
Frozen raspberries are a classic example of this problem.
The freezing speed of LN systems varies, as with air blast, with package
size but is typically only a few minutes, i.e. 20 times or more quicker than
standard air-blast systems. The speed of freezing is also important because it
enables a much smaller and cheaper set of equipment to process the same
throughput of frozen produce when compared to a conventional air-blast
system. Finally, the very quick initial cool-down of produce minimises
staling reactions and the chances for pathogenic bacteria or other spoilage
effects to grow.
A typical LN system comprises a cryogenic storage tank, usually rented by
the gas supplier, and a freezing tunnel. The tunnel has a moving stainless
steel mesh belt on which produce is placed. The product traverses first a cool
nitrogen gas zone, then a cold zone and finally is sprayed with LN from
sets of nozzles above the belt. Fans of moderate power within each zone
ensure a good convective thermal contact between and the produce. The
tunnel can be quite small, as small as 3 to 6 m long, as opposed to the 30 or
50 m lengths used in air blast, even though the belt is moving at a metre or
more per minute.
Advantages and disadvantages of freezing with LN. Although expensive in
running costs relative to mechanical air-blast freezing, LN is the best choice
in a substantial minority of applications, at least in well-developed
economies with low LN costs (Table 3.4).
Cryogenic air blast freezing. For very large installations used fairly
continuously, it is now possible to obtain the advantages of fast cryogenic
freezing by using expansion turbine air-blast systems. These operate a little
like a cryogenic air-separation unit, compressing and then cooling air back
to room temperature, drying the air and then expanding the compressed air
with a large and efficient t urbi ne, where it can cool adiabatically to 150C
(123K) or less.
The resultant ultracold air blast has many of the fast-freezing quality
advantages of liquid nitrogen freezing but avoids the use of liquid
refrigerants. The use of a heat exchanger on the exhaust from the freezing
tunnel or spiral means that wasted cooling power can be minimised. (A
spiral belt freezer is preferred because of its lower heat losses.) However,
the laws of thermodynamics predict that such systems will show significantly
higher power requirements than a conventional blast freezer; with Carnot
efficiencies, operating at 123 K will cost around 4.5 times as much power as
operating at 223 K. As well as this, the plant required is substantial and has
to be engineered for lower temperatures than are normal in the freezing
business so capital costs are expected to be high until systems of this type
become more common. By using the cooling power inherent in the air blast
right up to room temperature, the comparison can be made rather more
favourable (see next section).
LN belt freezing efficiency. To be efficient, nitrogen freezing must use not
just the latent heat of boiling of liquid but also the specific heat of cold
gaseous N
2
. This is achieved by employing the counter-current principle.
The specific heat of nitrogen gas from 77 K to room temperature (293 K) and
the latent heat of nitrogen as it boils amount to the following thermal
capacity. The theoretical heat absorbed per kilogram of LN at 1 bara up to
18C is:
specific heat (255-77) + latent heat = 1 (178) + 200 = 378 kJ kg
1
A figure reflecting achieved practice might be 300 kJ kg
1
. The mass of the
produce frozen (in water equivalent to 18C) per kilogram of LN is
heat absorption = water specific heat (293-273) + latent heat + ice
specific heat (273255) mass of produce
N
2
has a latent heat of vaporisation of 200 kJ and its gas specific heat
is close to 1 kJ so
mass of produce frozen 0.83 kg per kilogram of LN
If the cost of LN is $100/tonne, frozen food should cost about $0.12 per
kilogram more than fresh food; an acceptable premium for all but the lowest
value frozen produce. A comparative figure can be estimated for an air blast
freezer only if thermal efficiency and electricity prices can be assumed, along
with cost of capital for the much larger capital sum needed for the large
air-blast system and the real estate needed to house it. However, the overall
Figure 3.14 Principle of belt freezer using liquid nitrogen.
thermal efficiencies of LN versus air-blast freezing can easily be compared
on the assumption of ideal Carnot efficiency as follows. If freezing power
were delivered wholly at the cold-end temperature of the system
Energy needed (theoretical Carnot cycle) per joule cooling for air blast,
50C = 0.3 J
Energy needed (theoretical Carnot cycle) per joule cooling for LN, 195C
= 2.8 J
On the face of it air blast requires only 11% or so of the energy needed by
LN. However, this ignores the cooling effect of the sensible heat of LN and
the effect of this sensible heat being absorbed in the ASU plant. In effect,
only half the ASU plants cooling is at 195C; nearly half is at intermediate
temperatures. As a result of this consideration, a fairer comparison would
be that the air-blast technique only requires approximately 20% of the LN
energy input.
On this basis there would appear to be little to commend LN as a freezing
medium. Other factors, such as the possibility of freezing in a few minutes,
however, and the low capital costs of LN freezing mean that it is, in fact, the
most attractive option for many freezing plants. For very large plants one
option is to take advantage of the higher thermodynamic efficiencies of large
cryogenic plants. Very large freezing plants in a few locations have actually
employed an on-site LN plant. The 1970s operation of King Harrys Pizza in
Thame, UK, is an example.
The most common equipment for LN freezing is the straightforward
belt/tunnel freezer (Figure 3.14). Food is placed on a relatively slow-moving
stainless steel mesh belt. Occasionally a thin stainless steel foil is used which
presents a polished surface to the food being frozen. This can be useful to
avoid sticking and marking of the product base. Produce is normally taken
via zones of an intermediate cold until it reaches LN sprays and complete
freezing. In some belt freezers provision is made to achieve a rapid initial
freeze so that a frozen crust forms which will keep produce, such as tarts, ice
cream or piped cream, in shape until freezing is complete. Crust-freezing
can also be used to temporarily strengthen products which are not to be
totally frozen but are subsequently simply kept chilled or at ambient
temperature. The strengthened crust allows handling and package of the
produce to be achieved more easily with less damage.
Crust-freezing can also help with frozen or chilled prepared goods that are
distributed in vacuum display packs. Vacuum-packing is carried out by
wrapping goods on both sides with a rectangular piece of impermeable
plastic film, heat-sealing the plastic films together around three edges and
then, using the fourth side, putting a vacuum on the inside of the package,
removing the air and squeezing the product between the two films of plastic.
The crust-freezing of a prepared food such as pizza allows it to withstand the
crushing force of atmospheric pressure and avoid flash-off of moisture from
the surface when the vacuum is applied.
In some plants spiral-belt freezers are used. These have a continuous belt
onto which food is placed. It is then bent around into a spiral before being
brought to the exit again. They have a much smaller footprint on the
factory floor and have lower thermal losses, their compact shape having a
much lower surface area for a given volume. Strawberries, raspberries,
prawns, mushrooms and tomatoes are examples of produce commonly
LN-processed on spiral-belt freezers.
For more robust products, simpler and higher capacity systems can be
used for LN freezing. A simple cupboard can be used for batch freezing,
whilst for freezing of minced meats and similar diced products a slowly
rotating drum with screw-drive suffices and will achieve large throughputs in
a small space.
One important feature of frozen granulated or diced food is the
requirement that the frozen produce should not form large lumps but should
be formed as separately frozen individual pieces which are more convenient
for the purchaser. LN freezing techniques are more easily adapted for this
than conventional blast freezing, in that they achieve very high rates of
freezing, allowing crust-freezing to occur. Once this has happened further
cooling can occur without individual pieces sticking together.
For freezing of liquids into solid granules, a droplet freezer can be
employed. Although immersion droplet freezers are possible, a more
common system is to exude droplets of materials, such as fresh frozen egg
yolk (used in the catering trade), onto the outside of a rotating drum that is
cooled on the inside with LN. Individual frozen droplets are formed that are
scraped off the drum and packaged as free-flowing granules.
Immersion freezers. Before modern air-blast equipment became efficient
enough, freezing by immersion in ultracold (down to 40C) brine was
common. An immersion concept that has carried on into modern cryogenic
freezing practice is the use of an ice glaze to protect valuable and delicate
foodstuffs such as prawns. After initial freezing, produce is taken on the
moving belt briefly through a bath of water at a controlled low temperature.
Here it acquires a coating of ice, which is hard frozen by an additional
freezing stage after exiting the water bath.
Early attempts at ultrafast freezing relied on immersion in LN to achieve
the ultimate chill rate. However, there were often several practical problems
with this apparently simple approach and this method is still rather unusual,
although it has now become commercial practice in a few factories.
Freezing and chilling on the move. Once frozen, produce must be kept that
way. Similarly, there is a requirement to keep fresh produce at around 5C to
improve its keeping qualities. Large cold and frozen stores are now
common, some the size of a football pitch inside, with surprisingly small
refrigerator plants maintaining their interiors cold. A more difficult problem
is distribution of the produce while keeping it at a low temperature.
The most common solution adopted is the use of a separate diesel-driven
mechanical refrigerator plant on each semi-trailer or truck. However, these
units have some problems. If economically sized, to deal with the average
thermal loading, they are very slow to cool down after warming during
loading or unloading and the interior fluctuates in temperature consider-
ably. An alternative is to use a small tank of liquid nitrogen on the truck.
Using LN, almost instant cooldown and good subsequent controllability for
separate chilled and frozen compartments can be achieved in a truck or
semi-trailer unit.
Some high-quality fresh and frozen produce retailers have adopted this
ingenious LN-powered system for delivery to their shops. Goods are loaded
into the delivery truck, which is equipped with LN spray heads connected to
an onboard LN tank and control system. Once the doors are closed the LN
system comes into play, cooling the goods within a few minutes to the
required temperature. Moveable partitions between goods needing differ-
ent temperatures are available so that with a few additional components
trucks can deliver mixed loads of deep frozen, heavily refrigerated and
slightly cooled goods. An alarm is necessary to ensure that when the doors
are opened air mixes with the nitrogen atmosphere and an asphyxia accident
is avoided.
Freeze-grinding. For many flavouring foodstuffs, such as peppers, spices
or coffee, grinding is an essential part of the processing. Without grinding,
the products would not release their desirable flavours. However, grinding is
not easy for some products as they are not hard enough and they tend to clog
Figure 3.15 Ice adhesive shear strength for 10% glucose solution on a copper plate versus
copper temperature (courtesy of Air Products Corporation).
the grinding machinery. Worse, the grinding process tends to heat the
flavourings up, releasing some of the volatile flavouring essences which are
the materials raison dtre. The addition of a little liquid nitrogen to the
grinding machine avoids these problems.
The LN makes the foodstuff hard and less liable to clog, and keeps the
temperature so low that loss of essential flavouring oils and esters is
minimised. Nitrogen, in suitable systems, can provide an inert atmosphere,
which avoids the problem of oxidation that occurs particularly in grinding to
fine particle size. Fine particles in air can present a fire or explosion hazard,
which is eliminated if much of the air is displaced by nitrogen (see also
section 3.7.3). The LN cost can often be recovered in a premium price for
the product (as in freeze-ground coffee) or in increased throughput of
grinding machinery.
Non-stick freezing. By utilising to the full the ultrarapid freezing proper-
ties of liquid nitrogen, it is possible to freeze foodstuffs so fast that their
surfaces become non-sticky. Recent accurate measurements have con-
firmed a curious effect that the adhesion between a polymer surface and ice
formed on that surface is very strong when the surface is cooled to 20 or
30C but, surprisingly, is lower by a factor of up to 10 or so when the
surface is at 80C. Free-flowing minced products can be produced, as can
small fruits and small or sliced vegetables, without sticking. The latest
application of this ultrarapid freezing technique is in the production of
moulded fruit ice/ice-cream novelties; the Air Products Corporation offer
proprietary techniques in this area (Figure 3.15).
The science of why ultracold surfaces should offer such poor adhesion still
appears to be unknown. Although the phenomenon has been known for
some years in the preparation of frozen sections for microscopy using
LN-cooled copper surfaces, there is no comprehensive theory. It is thought
that ice formed on a cold surface bonds by first wetting the surface, then
freezing, hence the greater adhesion to wettable surfaces and the lower
(although still very substantial) bonding to non-wettable surfaces such as
PTFE. If a surface is not adequately wetted when it is very cold (a sort of
reverse Leidenfrost phenomenon), because the water freezes too rapidly,
then this could account for adhesion reduction. The other more obvious
phenomenon that comes into play is the difference in thermal contraction
between ice, water and the cold surface. Ice has double the thermal
contraction versus temperature of typical metals and shear forces can build
up at the ice/surface interface which, as the bulk of the ice forms, can cause
debonding after initial bonding has occurred.
The best guess about the reduction in adhesion strength at cryogenic
temperatures would appear to be a combination of freezing before wetting
and debonding following thermal contraction of the ice formed. There might
also be an unzipping or boundary crack effect, which is seen in other
adhesive bonds. Once a bond begins to peel off, stress is concentrated and
the debonding will propagate along the interface between two materials.
Perhaps thermal contraction effects can drive along such a debonding
phenomenon.
3.4.2 Dry-ice freezing
Where there are copious supplies of inexpensive carbon dioxide, this can be
a very reasonable cryogen for freezing food. It has the disadvantage relative
to LN that it can impart a taste to the food, particularly non-acidic food,
owing to its mildly acid reactive nature. Its use in delicately flavoured
non-acid produce is therefore sometimes not recommended. It is also not
capable of such a rapid freeze as LN, subliming at 194 K as opposed to the
77 K of LN. In addition, CO
2
at atmospheric pressure is a solid and cannot
form as intimate a contact with the food being frozen as LN does.
has a latent heat of vaporisation of 570 kJ (from the solid at
1 atmosphere but rather less from liquid, around 250300 kJ ,
depending on the pressure), as opposed to the effective 380 kJ kg
1
of LN.
This means that can be more economical in use than LN, provided it is
available at a low enough price. In the USA, the price in many regions is low,
so is used for refrigeration of freight cars on railways and truck trailers.
It is sometimes used in solid form as snow, which is loaded as liquid into
the roof of the freight compartment initially but allowed to flash-off in situ.
This is simpler than shovelling snow as a solid into the roof of the
compartment and also avoids the need for a control system for dispensing
There are some handling problems with In bulk applications, like
food freezing, it must be stored in refrigerated tanks. These typically hold
liquid at around 30 or 40C and approximately 10 bar. The tank
does not need to be an expensive vacuum-insulated vessel, although
insulation is needed. A modest refrigerator (a few kW) suffices for most
installations with a few tens of tonnes of stored. As noted above
(section 2.4.7), precautions should be taken so that the does not solidify
in the tank.
Liquid stored in this way does not contain the freezing capacity of
solid When it is sprayed out of a nozzle into 1 bar air, half of the liquid
sprayed immediately flashes-off as gaseous The resultant solid
snow can be mixed with a mixable product like minced beef or small
particulate foods, such as mushrooms, to increase contact between cryogen
and food. Freezing rates can be greatly increased by this. This amounts to
co-current heat exchange, however, and is less efficient than the effectively
counter-current techniques used in best LN practice.
The cooling of airline in-flight catering foodstuffs or portable ice-cream
sellers displays is carried out not by low density snow, however, but by
means of compressed bricks or slabs of solid of much higher density,
close to the 2 g theoretical density (section 3.7.2) made by compressing
snow with a hydraulic ram.
3.4.3 Gases as preservative chemicals
The use of gases for food preservation purposes other than cooling or
freezing food is now routine, although the volumes of gas used are much
smaller than in the freezing business. The preserving effects of gases such as
oxygen or carbon dioxide are modest but particularly valuable because they
are intrinsically safe and healthy, and do not affect food taste. Traditional
salt and sugar preservatives are unhealthy in excess and can affect taste
adversely, whilst many modern chemical stabilisers, although carefully
chosen and tested, will continue to have a question mark over their absolute
safety, at least in the minds of the general public.
The appearance and freshness of food in sealed plastic packages can be
enhanced by the correct selection of gas to be included in the package. For
example, oxygen will help to preserve the taste and the red colour of fresh
meat. Some further examples are shown in Table 3.5.
A lot of produce undergoes radical maturing changes over the period of
time after picking or other preparation and its consumption. Meat is usually
hung or kept for a few days, when it is subject to reactions which render it
less tough; game meats are too tough unless hung for a week or so.
Oxyhaemoglobin gives meat its red colour, but slowly gives up its oxygen
during hanging, giving the meat a livid blue/grey tinge. Oxygen helps
preserve the haemoglobin and hence the red colour a little longer. Fruits are
still to some extent living; they respire, taking in oxygen and giving off
Exclusion of oxygen from fruits can help restrain respiration and
lengthen storage, whilst avoiding the unwanted oxidation reactions (which
give rise to browning) which occur with imperfect or slightly bruised fruit.
The appropriate gas atmosphere will also help restrain pathogenic bacteria
and moulds in some cases; growth of Clostridium botulinum, for example, is
retarded in high levels of oxygen.
50
The storage and transit of fruit and other produce is often now enhanced
by systems which either seal in the appropriate atmosphere, as above, or, in
more sophisticated systems, actively manage the atmosphere by adding
appropriate amounts of nitrogen and
has been proposed for the preservation of milk but the proposition
has failed to find widespread acceptance. A milk with fizzes and has a
slightly acid taste rather than its usual more alkaline and creamy effect.
Consumer objections to the curious taste of the product might be expected in
simple liquid milk; too much magic to be acceptable would probably be the
advice of a food marketing expert. However, in milk destined for
cheese, butter, yoghurt and other dairy products, or as additions to
primarily non-dairy recipes, might be perfectly acceptable.
Ozone sterilisation offood equipment. The ready availability of inex-
pensive, electronically controlled ozonisers begs the question of whether
traditional liquid chemical sterilants such as bleach, sulphur dioxide
solutions or organic phenolic/resorcinolics/cresolics, used in the food
industry could advantageously be replaced to some extent by ozone. Ozone,
being gaseous, reaches all the internals of pipes and vessels more effectively
than a liquid, which may leave the tops of vessels and pipe loops ineffectively
sterilised. Furthermore it decays away to harmless oxygen, avoiding any
question of pollution, and is intrinsically safe and taste/odour free if any food
is accidentally contaminated.
Pumping. The mundane business of pumping beer or lemonade from a
tank or barrel to another tank, or to the customers glass, can be carried out
with a pump. However, it is often convenient, and more reliable, to employ
50
Carbon dioxide is not just used to preserve plant matter. can enhance the growth rate of
living plants (see section 3.4.5).
gas pressure to the pump. In essence this is the pressure transfer system
which is used in chemical works (section 3.3.8). The use of gas pressure is on
the increase, both on small and large scales. In large breweries, PSA
nitrogen plants are being installed for beer pumping. On a smaller scale,
although a gas cylinder is usual, small PSA plants, smaller than a gas
cylinder, are being considered for use in bars and pubs.
In the case of nitrogen, pressure transfer is used, i.e. gas pressure on
top of the liquid is used to transfer it from the storage tank to the point
of use. Carbon dioxide can also be used but it dissolves in drinks, a process
that has been pursued for centuries to enhance the taste. Carbon dioxide is
used in an ale-house to supply beer with not two but three effects : not
only does it pump the beer and add fizziness but the gas can be expanded
from the cylinder via a heat exchanger, or added to the beer in the line just
before the dispensing tap, in order to give a cooling effect. A 10C or more
drop in temperature is easily achieved, giving a pleasantly cool drink,
provided the gas used is sufficient and the cellar temperature is already fairly
low.
Sparging. Sparging is the injection of a gas into a stream of liquid to
remove an unwanted gas dissolved in that liquid. The most common
example is the removal of oxygen from water during the preparation of
beverages. Excessive oxygen leads to undesirable oxidation reactions in the
drink and limits shelf-life.
The process works because bubbles of nitrogen equilibriate with the
oxygen in the water until the nitrogen reaches around 10% or so oxygen
content. Thus, by blowing enough bubbles of pure nitrogen through the
water in a counter-current arrangement, most of the oxygen can be
removed.
Two patterns of sparging are in use: the in-line sparging bar and the
packed column sparger. The former is less effective but is cheap and very
small, whilst the latter effectively exploits the counter-current principle, and
can be made very efficient, but is high in capital cost and will typically need a
large space to put it in.
3.4.4 Carbonated drinks
Pressurised carbonated fizzy wines, such as champagne, and fizzy beers have
been made for at least 500 years. The fermentation process that produces
alchohol from sugar also produces carbon dioxide, which improves the taste
of the drink giving the beverage a fizzy acidic taste. However, a simple cork
seal in a glass bottle is incapable of retaining much overpressure of
limiting the fizziness factor to a slight tingling on the tongue. The practice of
Figure 3.16 Early Victorian lemonade bottle. The ball visible in the neck seals against the
narrower neck orifice to retain pressure. To drink the contents, the user pushes the ball
down. It is trapped by the crimp in the bottle from sinking to the bottom and prevented from
blocking the neck during pouring by the two lugs above the crimp.
adding has been employed since about 1800
51
although efficient and
inexpensive seals for the bottles for such beverages was a limiting factor for
many years (Figure 3.16). With the advent of effective seals for bottles, such
as the multilayer champagne cork and the internal glass bead check valve
seal, it became practical to retain a large amount of in a drink. Today,
common fizzy drinks, such as Coke, Pepsi or sparkling mineral water,
employ around 1% giving a pressure of a few barg in the worst case of a
bottle left in bright sunshine. One might expect that with this considerable
overpressure the opening of a lemonade bottle might invariably lead to an
instantaneous foam of drink and gas showering the unfortunate consumer.
This does not usually happen, however, because the is normally
released from solution rather slowly. If, however, nucleation centres are
51
Joseph Priestley, as well as discovering oxygen, made carbonated drinks and proposed in the
eighteenth century that the British Royal Navy could use fizzy drinks on long voyages with the
intention of reducing the vitamin-deficiency disease scurvy.
added to encourage bubble formation, or mechanical agitation applied, the
equilibrium is reached more quickly, resulting in a mass of foam.
52
Not all drinks require a high fizziness. Nevertheless, it is useful to have
some internal pressure of gas and exclude oxygen (to prevent undesirable
oxidation leading to rancidity or vinegar forming). The presence of a small
internal pressure of nitrogen enables the use of much thinner aluminium
cans for the canning of uncarbonated or low-carbonated beers and juices; a
practice called can rigidification. Special dosing systems which add just a
few drops of liquid nitrogen to an open can a fraction of second before its lid
is sealed on have been designed for can rigidification.
Guinness, the Irish black stout beer, is sometimes canned in special
nitrogen-pressurised containers; cans with a capsule in the base dubbed
the widget. The widgets function is to present the beer in a form as close as
possible to the real draught Guinness. The widget, containing nitrogen, is a
small chamber located in the bottom of the can which has a small hole in its
top. During filling, the widget becomes filled with beer nitrogen under
pressure. On opening the can and releasing the overpressure in the main
body of the can, the nitrogen is released and bubbles through the beer
causing nucleation of small bubbles of as well, the resultant very small
bubbles forming a satisfying fine frothy head on top of the beer.
There are still problems today with maintaining fizziness and excluding
oxygen over an extended period in plastic containers. Early plastic fizzy
drinks tended to lose via permeation through the plastic and
impermeable coatings were at first thought to be necessary. However, the
advent of the PET bottle, with its low permeation characteristic and high
strength enabled the production of thin, lightweight, inexpensive, high
capacity carbonated drink bottles.
Occasionally today, as in the past, lemonade bottles explode. The stresses
on bottles are quite high. The modern PET bottle is very thin and has to be
strengthened by bi-axial stretching of the polyester polymer to achieve the
necessary strength. The rupture of bottles can be due to a number of factors.
However, if the bottle was intact and correctly filled, then rupture is usually
due to excessive temperature rise in bottles exposed to full sunlight.
Interestingly, the most violent detonations are not from bottles which are
full but from those which are partly consumed. A little thought will show
that, as with hydraulic testing of gas cylinders, the energy release is greater
in the second case.
Ripening of fruit. Ethylene is used in small quantities during the transport
of bananas by ship. During their sea voyage, the bananas are subjected to a
small concentration of ethylene which helps to bring out the bananas
52
The reader is encouraged to observe the effect of adding finely milled salt to mineral water.
flavour and change the colour from the bright green of their fresh-picked
state to their familiar yellow colour.
The banana-ripening reaction using ethylene is essentially an enhance-
ment/speeding up of the natural process. It is catalytic or hormonal in action,
as can be judged from the effective amounts; a 50 g dose is sufficient to
induce ripening in tens of tonnes of bananas in an enclosed cold store.
Special small gas cylinders (ethylene tubes) are sold which contain around
this amount of gas.
The ripening effect of small amounts of ethylene is by no means confined
to bananas. Leaks at ethylene plants, although typically small and producing
sub-ppm concentrations of gas, have often been noticed to have effects on
local vegetation; for example, premature dropping of leaves from deciduous
shrubs and trees. Other fruits are also ripened by ethylene. Kiwi fruits are
very sensitive to ethylene and ripen more quickly in the presence of a few
parts per billion of ethylene.
As well as responding to ethylene, many fruits also evolve trace levels of
ethylene during ripening. As ethylene is evolved, it causes more ripening,
releasing more ethylene and so on. This autocatalytic action serves to
accelerate ripening and to ensure that ripening within a batch of fruit is
synchronised as the ethylene diffuses through the air in a store. Clearly, the
old saying it takes only one bad apple to make a barrel rotten is not just a
parable but actually contains a piece of science.
Ripening is delayed for many fruits if these traces of ethylene are removed
from the atmosphere surrounding the stored fruit. For some years now,
ethylene-removal systems sometimes as simple as a large bubbler with a
dilute aqueous reducing agent, for example, potassium permanganate, have
been used to reduce ethylene in large stores. Containers for shipping fruit
are now also being made which remove ethylene, although regenerated
absorber beds are favoured in these more high-technology devices. Shipping
containers, as well as removing ethylene, can also inject and/or nitrogen
using gas cylinders and PSA units to control the basic constituents of the
storage atmospheres. The BOC Group INTAC system is typical of the most
sophisticated of these systems, seeking to control ethylene and
water vapour levels by adding air, PSA nitrogen, cylinder or water,
whilst scrubbing ethylene or water, as well as governing the temperature
given by the refrigeration system.
3.4.5 Carbon dioxide in intensive agriculture
The basic action of a green plant is to convert water, sunshine, carbon
dioxide and a little nitrogen into more of that plants carbohydrates and
protein. In order to boost plant production, therefore, farmers have for
millenia added more water and sought to grow their crops in the sunniest
spots available, consistent with that water being available. It is only recently,
however, that farmers have sought to add the gaseous component, carbon
dioxide, in efforts to increase yield. (Small concentrations of another gas,
ammonia, have also occasionally been applied to intensive agriculture to
boost nitrogen.)
If a greenhouse can be more or less effectively sealed, it can be economic
to add to the atmosphere, boosting the natural 350 ppm to up to
1500 ppm. The acidulating effect of limits the amount that can be
added, which depends on crop type. Strawberries, cucumbers, table grapes
and tomatoes are all treated successfully with with yield per acre or
growth rate enhancements of the order of 20%. More controversially,
has been added to irrigation water with cotton and open-air tomato crops in
fairly arid areas. It is claimed that in alkaline soils acidulating the irrigation
water leads to enhanced dissolution of phosphorus and trace elements and
better crop growth. Although mineral acids can be used, has
advantages in cost and safety and, like mineral acids, helps to limit scaling on
irrigation pipework with alkaline waters.
Just as plants can benefit from carbon dioxide, brewing yeasts can benefit
from treatment with air enriched with oxygen,
53
whilst oxygen can be used to
enhance dough-making in the bread industry. Oxygen can be used to replace
the potassium bromate oxidising agent often added to bread, which is now
being banned in some countries as a potential carcinogen.
3.4.6 Gases in pest control
Pesticides are prepared in various forms but only a very few are actually
gaseous. The selective action usually required means that a complex
molecule, which is inevitably large and thus non-volatile, is indicated. The
use of liquid in high-pressure cylinders as a solvent for organic plant and
animal pesticides has now become a standard means of delivering these
materials. The liquid acts as a reasonable solvent for these materials,
although a co-solvent such as methanol or ethanol is often used. An example
of these pesticides is pyrethrum, the environmentally benign extract from
daisy flowers used for killing insects such as aphids. It dissolves easily in
liquid and is effectively aerosolised when the mixture is allowed to
spray out into the air.
Similarly, tea tree oil can be dissolved to quite large volumes in liquid
and such a mixture, on release, aerosolises the oil very effectively. Tea tree
oil is used to prevent the growth of bacteria and moulds inside apparatus; a
53
Oxygen is used by the yeast for its respiration during its normal growth. Yeast, like higher
animals, converts its food, sugars in this case, into energy by oxidation with producing
as a by-product. It is the which is relied on to give a beer its natural fizziness and to cause
bread to rise. It is only when deprived of oxygen that yeast starts to produce alcohol along with
the from the sugars it feeds on. However, the growth of yeast cell mass is enhanced by
typical use might be in bacteriostasis in air conditioning. (The Australia-
based Commonwealth Industrial Gases is developing the use of a tea tree oil
product in air-conditioning systems.)
Air-conditioning systems are known potential sources of bacteria and
other microbes. Damp parts of the system are particularly prone to
developing colonies of microorganisms, hence the outbreak of fatal
Legionella pneumophilia infections in buildings where air-conditioning
cooling-tower water became contaminated. Although air filters are in
general effective, old filters, particularly if they have become wet for some
reason, can let through numbers of organisms, which can actually grow
through filter material under some circumstances. Fortunately, the powerful
defence mechanisms of the human body, lysosomes in the nasal and tear
duct areas, for example, swiftly kill all but a few microorganisms. However,
some of the survivors will be pathogenic and may cause a disease directly or
exacerbate another condition such as asthma. Tea tree oil, dispersed into 10
or less droplets by aerosol flash-off, has been shown to be active
against Legionella and several other important bacteria.
Very few pesticides can be used directly on food for human consumption.
Even if they are not actually toxic to humans (as was arguably the case for
DDT, for example), their subsequent removal to low levels in the foodstuff
is usually required.
Simple carbon dioxide is one material that can safely be used. It is
harmless in any quantities in foodstuffs but can affect the growth and
respiration of fungi and plants. Hence can be used on its own for fungus
inhibition in grain silos. can also be used on its own as an insecticide in
grain silos. Because of its physiological effect on respiration at percentage
levels (it causes hyperventilation) it is effectively toxic in concentrations in
the 10% or so range and it is not necessary to achieve the 50% or more levels
needed if a simple inert asphyxiant gas such as nitrogen were used. Its low
toxicity in low concentration and its acceptability as a foodstuff ingredient
mean that there need be few fears about its safety.
Experiments in the US have shown that in grain treated with as little as
1 kg of per 200 kg of grain, insects are killed immediately following
hyperventilation and dehydration, whilst their eggs and pupae larvae are
also killed after a few days if the concentration of is high enough
(approximately 60%). Systems have been installed which blow into
grain silos at the top using a high initial flow rate for purging and then a lower
rate. The system then measures achieved gas-space concentration at
different points down the silo using a simple analyser (the sinks through
the grain).
A more potent agent for the destruction of insects in grain siloes is
phosphine. Although phosphine is highly toxic it is also highly reactive and
unstable, so it can be guaranteed that none will remain a few days after
treatment. Under normal conditions of moisture, the small amount of
phosphine used is swiftly hydrolysed and oxidised to a few parts per million
of harmless phosphoric acid
However, the high acute toxicity of phosphine means that stringent
precautions are necessary.
3.4.7 Margarine and edible oils
The margarine industry is said to have originated out of a published
government requirement for a substitute for butter during the economic
slump in Paris during the Napoleonic period when wars were exhausting the
normal means of production. However, animal fats have been substituted
for each other for time immemorial, so this could be viewed as simply the
first time that an attempt was made to make the substitute simulate butter.
Surprisingly perhaps, it was nearly 100 years later that chemical modifi-
cation of edible oils was carried out to extend this process of substitution.
The first practical process was patented in Germany in 1902 by Wilhelm
Normann, following the process still largely used today employing nickel
and hydrogen. This was only a useful process at that point, however,
because fairly pure hydrogen was becoming available from the burgeoning
coal gas and chemical industry. The fat-hardening industry expanded
rapidly, into baking fats and fats for soap makers as well as butter, and is
today a major consumer of merchant hydrogen gas.
The chemistry of hardening oils with Natural oils and fats are,
chemically speaking, esters, being the result of combining three carboxylic
(organic or fatty) acids with the trihydric alcohol, glycerol. The fatty acids
involved may be saturated (no C=C double bonds), unsaturated (one or
more C=C bonds) or polyunsaturated (two or more C=C bonds). An
example of an unsaturated vegetable oil is glyceryl linoleate (Figure 3.17).
Figure 3.17 Glyceryl linoleate.
In the presence of nickel catalysts, high-pressure hydrogen gas saturates
double bonds in oils, reducing them to harder (higher melting point) fats.
An example of a fat is glyceryl tristearate (Figure 3.18).
Figure 3.18 Glyceryl tristearate.
Many vegetable oils are unstable, reacting with oxygen in the air to form
rancid or off-flavours of fatty aldehydes and ketones, for example, as well as
discolouring. Hydrogenation generally reduces these problems by saturat-
ing some of the reactive double bonds both in the pure oil and its
contaminants, making the oil more stable.
Some polyunsaturated carboxylic acids, such as linoleic acid, are regarded
as essential components of the human diet. They are used by the human
body to make hormones and the like and are known as essential fatty acids
(EFAs). Saturated fatty acids, by contrast, are generally regarded as not just
inessential but possibly harmful, being implicated in coronary arterial
blockage heart disease. It is therefore an aim of most processes not to fully
saturate the linoleic and other polyunsaturates but to saturate only part of
the linoleic content and other acids whilst still achieving a higher melting
point.
There are problems in many oils with the linolenic acid content. Linolenic
acid has the formula Hydrogenated,
linolenic acid yields some valuable and some undesirable isomers of linoleic
acid. Some of these isomers cause a nauseous sweet odour to arise.
However, if unreacted, a substantial content of this acid will cause the oil to
polymerise. Linseed oil, which is mainly composed of linolenic acid, was up
until the 1930s the principal component of most oil paints as it reacts over a
few days in the air to form a tough skin.
There are many transition metal and noble metal catalysts that if finely
divided, will catalyse the hydrogenation. Copper has been extensively
researched and is valued for its removal of linolenic acid content whilst not
saturating all the linoleic content of an oil. However, problems with
poisoning and the necessity for virtually 100% removal of copper from the
final product have meant that nickel is still the mainstay catalyst.
Oil hardening in practice. The process is as follows. First, oil is prepurified
to remove air, particulates and catalyst poisons. It is then loaded into the
reactor vessel and vacuum applied, avoiding safety problems with
mixtures in the ullage space. The oil is then warmed to around 120180C
and agitated in the reactor vessel whilst hydrogen is bubbled through at
3 barg or so.
The nickel catalyst is one of the most ingenious aspects of the whole
hydrogenation process. Although fixed-bed catalysts are used, the most
popular system is the addition of finely divided (10 nm particle size) nickel
on kieselguhr silica support powder (c. to the reaction mixture,
followed by subsequent filtration. The filtration must be very effective; the
resultant mixture from the reactor is jet black in colour. The approach
usually adopted is to recycle the first few minutes of filtrate. The filter bed
accumulates caked catalyst, which acts as the true filter.
Commercial hydrogen purity is now well within the purity requirements
for alimentary fat hardening. Less pure sources of hydrogen are occasionally
used, particularly for less demanding applications such as soap feedstock
hydrogenation. Mercury contamination of electrolytic hydrogen is reduced
to infinitesimal levels by iodised carbon beds and oils made with electrolytic
hydrogen contain mercury levels much lower than many natural foodstuffs.
3.4.8 Decaffeinated coffee and detarred tobacco
Both nitrogen and carbon dioxide have had some impact in the business of
coffee drinks. Nitrogen is used both in grinding and in roasting while carbon
dioxide can be used for decaffeinating. Coffee beans ground with the
assistance of liquid nitrogen retain more of the flavour of the bean (section
3.4.1). Coffee roasted in a nitrogen atmosphere avoids the excessive
oxidation of some of the flavour compounds in coffee that can occur in air
roasting. Perhaps more ingenious than these applications, however, is the
use of carbon dioxide.
After roasting, coffee beans, and infusions made from them, contain
considerable amounts of the drug caffeine, which has both stimulant and
diuretic effects. A large market has been developed for a beverage with the
taste of coffee but without these side-effects. One process uses dichlorome-
thane as the caffeine solvent. This works tolerably well but has snags in the
cost and toxicity of dichloromethane, which must be very thoroughly
removed (an earlier process used benzene but this is now completely
unacceptable as even small amounts of benzene are considered carcino-
genic). Dicholoromethanes high volatility, it boils at 40C, makes this
relatively simple, although it is also highly inflammable on this account. A
more recent process using has many advantages over the dichlorome-
thane process.
supercritical extraction of caffeine. Supercritical carbon dioxide
(section 1.3.15) is capable of considerable solvent powers, including organic
substances such as caffeine, a fact which has been put to good use in the
processing of coffee beans. High-density supercritical is percolated
through the roasted coffee beans, dissolving a substantial proportion of the
caffeine. The caffeine is then removed in the and is absorbed in water or
carbon, allowing the to be recycled. The whole process is carried out at
high pressure, over 70 bar, which leads to the only economic problem in the
process: the cost of the pressure vessels required.
Tobacco tar-extraction and fluffing. Mass-produced tobacco products,
such as cigarettes, have now reached a pinnacle of sophistication. As well as
chemicals to keep the tobacco from going out and ultrathin clean-burning
paper, some gas science is also applied. In low-tar cigarettes, air is drawn in
through holes in the paper near to the filter end to dilute the inhaled smoke
stream. Only by laser drilling these holes accurately in each cigarette is it
possible to be sure that a consistent effect is achieved. Filter tips are now
carefully made too. Particular grades of triacetate fibre tow, compressed and
chopped up, are used. The fibres are parallel to the direction of flow and act
as diffusion filters, having little impact or sieving filter effect (section 2.4.3).
These offer reasonable retention of tar droplets and particles with low
pressure drop.
More sophisticated still is the use of supercritical fluid carbon dioxide to
process tobacco leaf prior to incorporation in cigarettes. Liquid
treatment of tobacco has several effects which tobacco companies value:
the removal of some harmful tars via its solvent effect;
the removal of the addictive nicotine via solvation;
residual left in the tobacco can be allowed to vaporise quickly,

leading to expansion of the tobacco leaf (fluffing).
The expanded tobacco is, of course, less dense than standard leaves and thus
less is needed to fill cigarettes, enabling still lower tar content ratings.
3.5 Gases in the semiconductor industry
The semiconductor industry continuously pioneers new processes for
processing materials. Every three years or so it doubles the capability of its
technology. Memory circuits, currently (1996) 4 or 16 Mbits in capacity, will
be 16 or 64 Mbits in capacity in less than six years. Processors, currently
running at 66 MHz will be running at 120 MHz with the same data in three
years time. The industry is on a technology treadmill in which it relies on
constantly convincing its customers that their current information tech-
nology, although only half-way through its life, needs replacement by
technology containing new semiconductors. The replacement cycle can only
be sustained by offering a substantial improvement, at least a factor of two,
at each approach to customers.
The semiconductor industrys need for constant technical improvements
and its high (albeit erratic) profitability, means that it is a hotbed for
innovation. Techniques used today for semiconductors are quite likely to
spin-off into other industries. Chemical vapour deposition (CVD), de-
veloped largely for the semiconductor industry, is now, for example, used
for optical fibre manufacture. The ultraclean gas techniques described in this
section seem particularly likely to spin off in the future into other industrial
sectors. For this reason, and because it is also a pioneer for processes using
gases, the semiconductor industry is particularly interesting and appropriate
for this book.
An integrated circuit (IC), is simply a large number of simultaneously
manufactured electronic circuits. Each individual circuit is made of conven-
tional electronic components, i.e. resistors (made of semiconductor),
capacitors (made of two metal plates insulated from each other with
insulating oxide), conductors (made of metal) and, of course, the active
components, i.e. the transistors used as amplifiers and switches (made of
semiconductor). There was a conceptual leap from circuits made from
laboriously hand-wired individual components to the IC in which conductor,
insulator and semiconductor materials are printed onto a substrate to form
the same circuit function. The IC, although it has a totally different scale and
physical appearance to the hand-wired circuit, has the same topology and
performs the same electrical functions. The real innovation of the IC is that
it can be mass-produced to a degree which dwarfs the productivity
improvements of organisational innovations such as the factory assembly
line. Productivity in terms of circuits produced per manhour has already
been multiplied a million-fold.
The IC plant or wafer fabrication plant, often shortened to wafer fab or
just fab essentially prints (fabricates) active microelectronic circuit layers
onto thin round wafers of silicon semiconductor, 0.5 mm in thickness and
mostly 100200 mm diameter.
54
The single crystal silicon substrate, or a
silicon layer contiguous with it, is used for fabricating the transistors, whilst
upper deposited silicon layers can be used for resistors. Metal layers are used
for conductive interconnections, whilst insulating layers provide isolation
from conductors and the dielectric portion of capacitors. The patterning for
each layer is defined by UV exposure of a mask pattern onto a photosen-
sitive plastic resist layer applied to the silicon wafer. The resist is
polymerised only in the exposed areas and the unexposed resist removed in
the development process. Other processes can then be used, such as etching,
which will only affect the exposed wafer, whilst leaving the resist-protected
parts unaffected. This simple principle can be used up to 20 or 30 times (20 or
30 mask layers) to form the necessary circuit topology.
54
The term wafer came about because originally the largest single crystals of semiconductor
that could be obtained were only a few millimetres or so in diameter and the 0.5 mm slices from
these used for early integrated circuits were indeed very similar to a Christian communion
wafer. With fut ure fabs pl anni ng to use 300 mm diameter silicon, a new term is perhaps called
for.
Figure 3.19 An integrated circuit with diode isolation; (a) electronic circuit; (b) integrated
circuit layout.
Throughout a wafer fab, industrial gases are taken for granted as an
integral part of nearly every procedure carried out. It is often the case that
when all the major applications for gases inside a wafer fab are added up,
they total much less than the gas consumed by the site. This reflects that in
fact there are a myriad of minor applications which together consume large
volumes of gas. Semiconductor manufacture is arguably more dependent on
industrial gases than any other manufacturing process today.
3.5.1 The need for high purity in semiconductors
Gases are used in many parts of the processes in the manufacture of
semiconductor integrated circuits. Nitrogen is ubiquitous in the wafer fab. It
is used for all sorts of processes, from drying wet silicon wafers to diluting
process gas. Other bulk gases used are oxygen, argon and hydrogen, in much
smaller amounts. Only cylinder quantities are needed of most of the
hundred or so other gases used but these are the essential reactive gases,
from the silane used for epitaxy and glass insulator deposition to chlorine-
and fluorine-based etching mixtures and even more exotic compounds used
in quantities of a few grams for doping.
The requirements for gases for making semiconductor devices are
distinguished from most other applications for gases in that ultimate purities
are generally needed. The silicon on which most ICs are made is itself an
ultrahigh purity product. The 150 or 200 mm diameter silicon wafers
processed in wafer fabs today are thin slices from giant single crystals, the
biggest commercially used monolithic crystals, and are pure at the level of 10
parts per billion.
The wafer fab is an ultrahigh purity manufacturing environment which is
cleaner by far than, for example, a hospital operating theatre. Special
non-dust-generating cleanroom bunny suits and face masks are required in
the cleanroom. An ordinary room environment has millions of tiny particles
of dust per cubic foot whilst in the cleanroom there are fewer than ten
particles; less than a bacterium in each cubic foot of air. Particulates as small
as virus particles, are effectively monitored and controlled in the
latest cleanrooms. Still higher levels of cleanliness are achieved near the
silicon wafers themselves by the use of mini-environments. These are the
suites of equipment now used where wafers do not contact even the
ultraclean cleanroom air but only see air or nitrogen inside the equipment or
inside the SMIF (standard mechanical interface) boxes used to transport
wafers between different parts of the cleanroom.
Many of the processes in the wafer fab have a special purity requirement,
as a glance down the list in Table 3.6 will confirm. Ultraclean gas technology
has therefore grown up largely to meet semiconductor needs. Only in a few
cases is there is good quantitative evidence for the effect of impurities on a
semiconductor process. It is well known, for example, that oxygen and
moisture impurities in gases used for CVD silicon nitride growth give a
predictable proportion of oxygen in the nitride layer grown (CVD nitride is
often used to passivate circuits, protecting them from reaction with the
atmosphere). However, in many other processes theoretical predictions are
clear-cut or qualitative results are available so that a picture has been built
up, albeit incomplete, of the most critical impurities (Table 3.6).
Particulate contamination. A particle of larger than may ruin an
integrated circuit using design rules, as is representative of the
thickness of some these IC structures. It is possible for even smaller particles
to cause problems if they are composed of poisons such as alkali metals.
The latter can diffuse freely through the semiconductor structure, poisoning
a volume of silicon much larger than their own volume. They cause
unpredictable and time-varying doping, off-spec device parameters or,
worse, drift of device characteristics over time.
For any particular scale of circuit, the maximum allowable particle size
can probably be set at 10% of design rule for ordinary particulates and
perhaps between 1% and 3% of design rule for poisons. As far as the
number of particles above this limit that are allowed, a typical standard calls
for less than 10 or less than 100 per cubic foot. As this is in excess of what
could be typically guaranteed for industrial gases in bulk, most semicon-
ductor processes will require filters on all critical processes.
Particulates are of little or no importance once the wafers have been
passivated as the active devices are then largely protected against any
foreign body intrusion. Higher levels of particulates can also, perhaps
surprisingly, be tolerated in some of the vacuum processes, such as
low-pressure CVD, plasma etching or sputtering. This is because at the low
pressure of operation the particles can settle, just like sand in air, and
because the processes are frequently performed with wafers upside down or
sideways, reducing the possibility of particles settling on active surfaces. In
addition, some processes are always followed by a wet chemical cleaning
step that will also remove particles. (CVD and etching can create large
numbers of particles themselves.)
Reduction of particulates is not easily achieved by clean-up after particles
have been deposited. This is because the electrostatic forces developed
between the wafer and the particle are very large compared to the particle
weight. If the force between a spherical particle and a conducting surface is
calculated on the basis that it adheres with an electrostatic force (due to a
small charge on the particle), for example, is easy to show that the force of
adherence, expressed in terms of the gravitational strength required to
dislodge it, varies as where R is the particle diameter
particle weight,
contact electrostatic force,
electrostatic force in gravities
A similar effect, i.e. very high adherence forces for small particles, is also
seen if the weaker van der Waals forces are assumed between particle and
surface. Given the difficulty of removing particles once deposited, it is
important to avoid deposition in the first place. A reduction in particles
derived from gases can be achieved by filtering gas before the point of use.
However, care must be taken that the filters used do not add particles
derived from their own structure to the gas stream. This is a typical
characteristic of, for example, fibrous filters.
Removing particulates from wafers usually requires a wet process. At
entry to the wafer fab, wafers are often wet-scrubbed with rotary polymer
sponge brushes (which removes particles). Wet processes are good at
removing particulates because a liquid, by acting as a dielectric between
particle and surface, can reduce the adhesion forces. In addition, a liquid can
coat the particle and cause it to be absorbed into the body of the liquid as a
colloidal particle. At various points in its progress through the wafer fab,
therefore, wafers are subjected to wet cleans, though not with brushes after
the initial scrubbing. Wet cleans often employ an acid/oxidising bath and an
alkali/oxidising bath (section 3.5.10). Oxidation can help remove particles
by oxidising the particle and/or the wafer surface at the point where the two
are adhering, enabling the particle to be detached.
Gaseous contamination. Generally, the standard industrial gas specifi-
cation, in the low parts per million for most impurities, for bulk gases is
adequate for much semiconductor use as far as dissolved impurities go.
However, there are many critical applications where this is not the case.
Silicon and metals of ICs will react with even tiny amounts of water and
oxygen in high-temperature inert gas processes such as annealing. A typical
wafer fab plant may require these amounts to be below, say, 50 ppb and
10 ppb respectively, for all its high-temperature argon and nitrogen
processes and these specs are constantly being forced to higher purities.
Even greater purities are desired for the silane, hydrogen, nitrogen or
ammonia used in nitriding. These must contain sub-ppb oxygen as silicon
dioxide is thermodynamically favoured over silicon nitride and the latter will
only form stoichiometrically in the complete absence of oxygen. Frequently,
layers are seen which correspond to silicon oxynitride caused by very slight
oxygen contamination during the reaction.
For many processes the common impurities in oxygen, water and
hydrocarbons, are regarded as unimportant. However, some companies
have insisted on low hydrocarbon content for Mbit and above circuits and a
number use purifiers to lower the water content to the 1 ppb region. These
are mostly using a dry oxidation process for the critical gate oxide, which
small quantities of water can render uncontrollable by accelerating oxide
growth in an unpredictable way. Carbon contamination can cause gate oxide
problems.
It is sometimes assumed that purge gases can be of low quality. This is not
the case. HCl and the other corrosives etchants will react with moisture and
attack metal parts, producing metallic contamination if the purge gas
contains moisture. Silane and the other pyrophoric gases will react with
residual oxygen to produce oxide particulate contaminants. Purge gases in
general must be of similar quality to those used in the process if they are to be
effective in maintaining the gas lines of the critical processes.
Table 3.7 lists the level of purity now achieved (and expected) in state of
the art wafer fabs. All these figures should only be taken as example
achieved purities. Plant and purifier performance vary widely with respect to
these minute trace levels of contaminants. Note that the moisture figures are
probably high because of the use of Teflon polymer filters. At the moment
metal or ceramic filters are still exceptional but when these become
widespread these figures may well fall to the 0.1 ppb or better achieved in
ultraclean laboratories. At 0.1ppb there is a consensus that bulk gas purities
are probably good enough for 99.9% of all wafer fab operations and that
emphasis should now shift to purity of vacuum, surfaces and special or
process gases.
3.5.2 Making ultraclean gases
Ultrapure gases, it turns out, are not in general too difficult to produce.
Physical distillation increases purity in principle by geometric steps. After
each distillation the gas contaminants decrease by the same multiplying
factor. If one pass up the purifying column, or one distillation plate,
decreases contaminants by a factor of 10, then two columns or plates will
increase contaminants by a factor of 100. Ultrapure grades of most gases can
be created by physical separation; air gases can simply be drawn off the
standard liquid air column and then redistilled in a small-diameter side
column. Gases such as silanes are also fairly simple to distil as part of the
production process.
It is now generally accepted that nitrogen generation on-site is a more
economical way of generating nitrogen at ppb levels of purity. By adding
distillation trays and small modifications to ancillary parts of the plant
(adding extra beds to the air prepurifier units, for example) levels of around
1ppb can be achieved with the exception of CO, which is not effectively
separated in air distillation. Depending on plant siting (CO mostly derives
from vehicle exhaust in urban areas), this could be as high as a few hundred
ppb CO with occasional peaks up to ppm level in some sites. The addition of
a purifier(s) is justified if ppb or lower levels of CO are needed.
Chemical purifiers are capable of even more spectacular decreases in
contaminants in one pass, even more so if they can use more or less
irreversible chemical reactions (section 2.2.11). Purifiers offer the conveni-
ence that it is only necessary to purify gases used in critical processes and
offer an insurance policy in case a batch of off-specification gas comes
through.
With gases produced in bulk by the general chemical industry, ammonia
or hydrogen chloride, for example, there is more of a problem. After
attempting to purchase the best grade of gas in the best container available,
the gas company supplying the semiconductor industry may be able to select
batches. Batches not up to semiconductor standard can be sold for less
critical applications. Failing this, there is no alternative but to build a special
purification rig. This may be as simple as a blowdown facility to remove
volatile gaseous impurities such as oxygen on top of a liquefied gas such as
HCl. Similarly, vaporising and filtering the product, leaving heavier
particulates at the bottom of the cylinder, can also be carried out. A
chemical absorber rig can also be very simple, comprising an absorber
material on inert packing. The chemical chosen needs to match both the
substrate gas and the impurity gas, however. Ultimately, if none of these
simple solutions suffices, a distillation rig for upgrading gases is used.
One elegant solution for distillation purification employs the cryogenic
storage tank itself as a still (Figure 3.20). Liquid hydrogen tanks are
available which allow vapour withdrawal, the lost heat of vaporisation being
added back to the tank via an external ambient heat exchanger on the
withdrawn gas and an internal coil heat exchanger in the tank. A regulator
ensures that tank pressure is not raised too far, when direct vapour
withdrawal without back-addition of heat occurs, until the tank pressure has
fallen far enough. In the bottom of the tank, solid contaminant gases can
build up; with a hydrogen/oxygen slurry a few grams could explode and
rupture the tank if ignited. A periodic withdrawal, perhaps only every few
months or so, of LH from the bottom of the tank avoids the problem. This
Figure 3.20 Principle of the self-distilling liquid hydrogen tank.
system works well for liquid hydrogen, partly because the heat required to
boil is so small that a small heat exchanger in the tank suffices. Perhaps,
despite problems with the larger heat flows needed, self-distilling tanks will
be devised for other gases in the future.
3.5.3 Keeping ultraclean gases ultraclean
Keeping gases clean once made is as big a problem in ultraclean technology
as making them in the first place. Basic ultraclean pipework and vessels are
now made as much as possible from 316L stainless steel, often electropo-
lished. The 316L grade of stainless steel has 16% Cr and 10% Ni. It is
completely rust proof under normal conditions and highly resistant to most
chemical attack, even chlorine under dry conditions. Although copper
tubing is still encountered in wafer fabs, and can give perfectly acceptable in
service with more benign gases, it is no longer used in new installations.
Copper tubing takes longer to dry than electropolished stainless and is
subject to corrosion if accidentally exposed to contamination. Mill-finish (or
as-drawn, i.e. not polished) stainless steel still has its devotees, however.
Although mill-finish tubing does take longer to dry-down to low moisture
levels, at least when sub-ppb levels are needed, the presence of polymer
filters or process chambers which contain moisture anyway may mean that
this is not important. In fact maintaining continous flow in a gas line may be
just as important as the tubing finish, since tubing with little or no flow will
always show higher levels of contaminants than tubing with a good flow.
Figure 3.21 Electropolishing of stainless steel tubing showing polishing electrode inside tubes
along which is flowing the electropolishing solution (courtesy of Accles & Pollock Ltd).
Electropolishing. The electropolishing process, as the name suggests,
gives a mirror-like polished surface to the stainless steel. Electropolishing
also has the effect of increasing the surface concentration of chromium and
especially nickel, which are the elements that lend stainless steel its thin but
adherent and unreactive oxide surface. Electropolishing is carried out by
making the surface to be polished an electrode in a proprietary aqueous
solution of salts, inserting an inert electrode, then applying an electrolysis
current. More current flows from sharp points on the stainless steel surface
than from the lands in between. In this way, high spots are removed and the
polishing action is achieved. At the same time, iron atoms are removed more
rapidly than chromium or nickel atoms, which enrich the Cr/Ni concen-
tration. A carefully electropolished surface has a much smaller tendency to
trap particulates than the standard stainless steel smooth finish and it also
retains much smaller amounts of moisture.
Passivation. Fluorine passivation is used in installations for corrosive
fluorine-containing gases and has been used on other gas supplies, although
there do not seem to be many numbers available on its performance. It is
carried out using a gas mix with a small amount of fluorine in it, which forms
a thin but stable fluoride film in a slow deliberate manner. Subsequently, the
gas line will be safer to use with fluorine or HF gas and will give lower levels
of contamination.
Stainless steels of the standard 316L variety are normally passivated by
passing a mixture of at least 2% in down the pipework at ambient
temperature. A fluoride layer about 6 nm thick forms, basically overlying
the pre-existing native oxide, which is 2 or 3 nm thick. This fluoride layer,
mainly chromium and iron fluorides, is passive. It will not react further with
fluorine nor with agressive fluorides such as (which otherwise tend to
react, causing corrosion and plating out on the inside of the pipework).
Presumably the presence of involatile iron fluorides stops migration of metal
ions to the surface, preventing further reaction.
Unfortunately, standard fluorine-passivated surfaces are spoilt by ex-
posure to ambient moisture and a passivation procedure must be repeated
each time a gas distribution line is exposed to air or other moisture-bearing
gas. Professor Ohmis team at Tohoku University, Japan, have published an
improvement to fluorine passivation, which relies on exposing pipeline
components to 100% fluorine at 220C for an hour followed by 320C
annealing in nitrogen (Ohmi et al., 1988). This process is clearly much more
hazardous than 2% treatment and cannot be done in situ because of the
need for strong heating. However, this treatment produces a thicker
passivation film and one that is resistant to moisture attack (see also section
3.5.5).
Pipework joints. Electropolishing, high-grade stainless steels and passi-
vation are all capable of improving gas quality for a wafer fab but there are
problems. The first problem is in jointing stainless steel. Plates can only
effectively be joined by welding and this is also the most common means of
joining pipes. Pipes are conveniently joined with the automatic orbital TIG
welders which are now quite common in the industry. The choice of the low
carbon 316L grade of stainless steel is explained here. The low carbon means
that TIG or MIG welding will lead to a HAZ with little unwanted hardness
and a good quality joint. Unfortunately, welding does lead to the generation
of fine fumes of submicron metal particles, though this can be controlled to
some extent by inerting with argon inside the vessel or pipe as well as
outside.
It is obviously not possible to produce an all-welded system for gases; it is
always necessary to have demountable joints. It is also convenient not to
have to weld in confined areas such as in small underfloor ductwork and
inside equipment. Although many systems still use Swagelok-type crimped
ferrule compression joints (section 2.4.1), standard practice is now
type demountable joints.
Good though they are, Swagelok type joints have small narrow crevices
for particles to accumulate in, and expose part of the outside of the pipe to
the gas contained. A better mechanical joint for applications needing the
ultimate in cleanliness and gas purity, such as wafer fab plant gases, is a
face-seal joint such as the VCR. These special end fittings, which are welded
onto tube ends or machined onto components, have a face with a
ring-shaped ridge. These annular ridges cut into a flat soft washer placed
between the two faces as the union nut clamps the two ends together. The
washer is replaced each time the joint is put together. The joints are similar
in many ways to the all-metal joints used in vacuum work where knife-edge
rings on the faces are forced to seal on a soft copper washer. These fittings
are more complex to apply than Swagelok fittings because of the require-
ment to weld the fittings on either end of a tube, so they are only used when
ultraclean gas is definitely required.
MCG face-seal fittings take the principle of face-sealing one stage further
by adding a ball-bearing to the union nut so that there is virtually no
tendency for the pipe to rotate during the assembly of a joint and a more
highly leak-tight and contamination-free joint can be guaranteed, giving a
welcome bonus of a joint that needs only minimal force to tighten it. MCG
fittings are used with C-ring hollow metal seals, giving improved freedom
from particulate contamination.
Ultraclean valves. The next weak point which must be attacked in an
ultrapure gas installation is where some kind of flow control fitting is fitted.
A conventional control valve, with its packed stem, its soft polymer seat,
springs and other metal parts not made in the same 316L stainless steel, is
clearly not going to be satisfactory. The polymer seat may outgas moisture
and oxygen, the packing may leak air, as well as being a source of
contamination itself, and the sealing surfaces will in general release particles
on operating the valve, especially if the valve is neither fully open nor fully
closed but in a throttling condition.
An ultraclean valve should ideally be all metal with a hard metal valve seat
and a stainless steel diaphragm or bellows to separate stem and spring from
the gas stream (Figure 3.22).
Other special features may be needed in control elements. For example,
the use of tied-diaphragm regulators. Instead of relying on the inlet gas
pressure to push the valve stem onto the diaphragm, these have the stem
welded to the diaphragm. This avoids erratic operation when small particles
form, for example, when silane or contacts residual oxygen or
moisture. The oxygen or moisture produces particulate silica or borax in the
system which can get on the valve seat and prevent correct operation.
Ultraclean gas control panels. Ultraclean gas control panels and ultraclean
gas cabinets for toxic or corrosive cylinder gases follow the same principles
as standard equipment (and indeed, some standard equipment now follows
Figure 3.22 Pressure regulator for ultrapure gases. Fabricated in 316L stainless steel with
electropolishing on inner surfaces and a fluoropolymer seat material. The diaphragm is tied to
the valve to give a positive shut-off for toxic gases (courtesy of Hale Hamilton (Valves) Ltd).
semiconductor industry practice). However, apart from the selection of
ultraclean-compatible components such as diaphragm valves and perhaps
electropolished parts, there are a number of features which differentiate
them:
they consist of an all stainless steel (316L) largely welded assembly;
the lengths of connections between components are optimised to

minimise lengths of pipe that will be dead-volumes in operation;
access to vacuum source for the pressure purge cycle is essential (ideally a
rotary vacuum pump at a fraction of a millibar pressure);
automatic purge operation is usual;
they can be remotely controlled by computer (with local override) for

complex installations via pneumatically operated diaphragm actuators on
many or all valves;
they have electronic monitoring of cylinder contents (weighscales or

pressure);
electronic pressure gauges are used to avoid the dead-volume of a

Bourdon tube gauge;
a relief valve is usually necessary after the regulator to ensure that

Figure 3.23 Manually operated ultraclean gas cabinet using electropolished VCR components
and vacuum-assisted purging, designed for one-purge cylinder (right) and two-process gas
cylinders (courtesy of Cambridge Fluid Systems).
sensitive downstream equipment cannot be overpressurised in the event
of a regulator failure, as well as for safety reasons;
an excess flow valve or orifice is desirable, again to protect downstream

equipment as well as for safety reasons;
check valves are required to prevent back-contamination of purge gas;
a separate cylinder of gas for purging is needed to totally eliminate the

possibility of back-flow of process gases into the site main gas supplies;
a pneumatically-operated cylinder valve is desirable for safety reasons;
the whole assembly is put together in cleanroom conditions.

Note that both the connection to the cylinder as well as downstream of the
regulator, can be purged in the gas panel shown in Figure 3.23. The purge
cycle may be automatically repeated up to 10 times, which is why computer
control is very convenient, releasing the operator from this tedious job and
reducing the possibility of error.
Ultrapure gases in wafer fab equipment. Much effort goes into achieving
ultraclean gases up to the gas suppliers definition of point-of-use, i.e.
where the pipe carrying the gas enters a piece of equipment or the process
chamber where the silicon wafer will react with the gas. However, the
conduction of the gas to this point in ultraclean condition is futile, of course,
if an ultraclean reactor is not then used to process the actual wafer. The gas
must be kept ultraclean right up to the point-of-reaction (POR).
In well-built facilities it is not so much in the gas pipework or storage
system but more often closer to the POR, inside the process reactor, that
many wafer fab cleanliness problems seem to occur. When apparently
ultraclean reactors are opened up, rubber O-ring seals (potential outgass-
ers), plastic pipes (these can diffuse moisture and oxygen into the system),
cracks in glassware and leaking valves are often found. The latest generation
of semiconductor equipment does seem to have eliminated most of these
weak points when new, although this equipment will still be subject to
degradation via cracking and valve failure with time.
Some of these problems are difficult to avoid. A recent short study from
the ultraclean specialist research group of Marc Heyns of IMEC, Belgium,
for example, showed flaking from the inside of a chamber used for Ti
sputtering. Just 250 ppb of in the nitrogen or argon gas admitted to the
chamber had led to the formation of the oxide, which caused particulate
problems in the process.
55
There are two general trends in wafer fab equipment which will tend to
improve gas purity at the POR. There is a trend towards performing wafer
processing in cluster tools, i.e. several related operations are performed on
a wafer without that wafer being removed. The wafer is simply shunted
around by a robot, operating in a vacuum, from one process module to
another. A cluster tool with four chambers for etching might, for example,
carry out a tungsten etch in followed by an aluminium etch in then
a titanium/tungsten barrier layer etch in HBr, followed by a
photoresist stripping operation in oxygen plasma, before the wafer is
returned to the cleanroom environment.
55
Other problems may seem so huge as to be easy to avoid. The author has measured up to 20 or
30 ppm during the ramp-up of a rapid thermal processors, although once up to temperature
this level could be swiftly purged out down to ppb levels again. Similar levels of were
seen in rapid thermal processor experiments recently during the testing of a fast moisture
analysis technique. However, high levels of moisture in this case may well relate to moisture
taken into the reactor absorbed onto the surface of the wafer. A monolayer of moisture would
give moles on a 200 mm wafer, for example, enough to give 10 ppm for a minute or so at 51
per minute purge gas flow.
The other important new ultraclean processing development is the move
towards mini-environment processing. In this, each piece of processing
equipment, whether ion implanter, oxidation furnace or plasma etcher, has
its own filtration system for clean air (or nitrogen) and is otherwise sealed off
from the cleanroom except for its SMIF port. Wafers to be processed are
brought in the SMIF sealed box to the port, where the machine can unload
the wafers in an loading airlock (loadlock) and then move them into the
machine to be processed; the wafers will not contact cleanroom air. Newer
SMIF-type systems such as the INCAM (individual cassette manufacturing)
system, from the CEA-LETI group in Grenoble, France, may eventually
take over from SMIF to achieve even higher levels of performance, even
with the huge 300 mm wafers expected soon. INCAM is being designed to
handle single wafers or wafers in batches of five or so.
3.5.4 Ultrapure gas pipework simulation
There is little doubt that the best practice for ultraclean pipework layout can
be encapsulated in two principles: avoid dead-legs and ensure continuous
flow. Within these broad principles, however, there are many detailed
compromises to be made and computer modelling can help this process.
The behaviour of trace contaminants in a gas system is not readily
predictable, especially when the system is complex. A relatively simple
computer model, simulating the effect of dead volumes, dead or low flow
pipework legs, under different assumptions of flow rate, input purity and
leak rates, can be very useful in pointing to the parts of a system critical to
purity. It can be useful both in the original design of systems and in the
commissioning and improvement of existing pipework. The simulation of
the effects of moisture retention, although less tractable analytically, is also
useful. Ideally, such a system works both for steady flow conditions and
during transients (e.g. start-up or gas-line switching).
Such a pipework simulator was first put under development and offered to
semiconductor users by the LAir Liquide company, although many of the
other gas companies have now followed suit. The models have the following
features:
modelling of dead volumes of different shape;
modelling of contaminated dead branches;
modelling of different (but constant) outgassing rates.

They require the input of a pipework connectivity and flow map, and output
a predicted purity map. These inputs are then applied to a model of the gas
volumes and absorbing surfaces. The latter are particularly important for
predicting moisture. Moisture on stainless steel surfaces follows an absorp-
tion principle a little different from the Langmuir isotherm (described in
section 2.1.6) for absorption on PSA zeolite absorbers. Rather than having
many sites for absorption that have the same binding or absorption energy,
moisture on stainless steel is absorbed strongly (high binding energy) for
small surface coverage and weakly (low binding energy) at high surface
coverage. This follows from the irregular polycrystalline nature of the steels
surface where there will be a variety of different sites, of varying binding
energy, the high binding energy ones (more stable) being filled by moisture
molecules first. With this kind of isotherm and algorithms linking volumes,
surfaces and flows together, fairly precise modelling of pipework behaviour
can be achieved.
There are other phenomena which impinge on purity achieved in
pipework systems, i.e. back-diffusion and leaking into the pipeline.
Back-diffusion (see also section 1.3.10) occurs where impurities from a
process chamber leak back into a pipe supplying them with an ultrapure gas
against the stream of gas issuing from the leak. The most sensitive method of
leak detection available should always be employed in the fabrication of
semiconductor gas lines. Generally, gas lines and most vessels can be
evacuated, allowing the use of the mass-spectrometer/helium technique
described in section 2.4.4.
3.5.5 Further developments in ultraclean gas technology
In addition to the basic ultraclean technology outlined above, gas tech-
nology for semiconductor gases has come to encompass hundreds of
technical improvements, of which the following are some of the more
important.
Heated filters. When a new PTFE-based filter is installed in a gas line, the
gas flowing down the line will typically be contaminated to a ppm level for a
few hours, depending on the size and type of filter, and the gas flow rate, by
outgassing from the huge surface area of the filter medium. It will still have
ppb contamination due to the new filter after days of operation. Many wafer
fabs expect to use heated filters in the future so that outgassing can be carried
out after installation or replacement. With conventional PTFE filters,
however, heating can only easily be achieved by heating the purge gas used
and the maximum usable temperature is only approximately 200C as the
PTFE-based materials will degrade at 250C.
Metal and ceramic filters. The Teflon membrane filter is the mainstay of the
semiconductor industry filter. It is relatively inexpensive, very efficient in
filtration and chemically highly resistant. However, it has the considerable
disadvantage that it can absorb moisture and other contaminants from the
atmosphere, slowly releasing these into the gas stream being filtered.
Advances in sintering technology mean that alumina-based ceramic and
metals such as stainless steel or nickel can now be prepared as sintered
blocks offering fine pore size and good filtration properties. The latest fil-
ters are based on technology developed for ceramic membranes for use in
uranium hexafluoride diffusion isotope separation (supplied by, among
others US Filter and SCT). A strong (2 or 3 mm or more) support of rather
macroporous (20 m and larger) ceramic alumina is coated with several
layers of a fine alumina slip ceramic that, when fired, forms a thin layer of
microporous ceramic with pore size from 10 m down to a small fraction of
a micron and just a few tens of microns thick. Several layers of slip are used
so that pinholes can be avoided and a gradated transition from coarse to
fine texture managed. In recent filters, filters with three layers of success-
ively finer alumina are deposited. The result is a robust membrane filter
with a texture of the order of 0.1 m that can efficiently filter gases to a
very high standard without blocking or blinding. Gas flow is into the
macroporous alumina, through the microporous layers and out on to the
clean side. Nickel and ceramic filters have the advantage that they can be
used with highly reactive gases such as HC1, HBr, and dichlorosilane.
Sintered filters typically have to be 10 or 20 times thicker than equivalent
PTFE membrane filters. The most penetrating size (section 2.4.3) for a
PTFE filter is around 0.04 m with a theoretical penetration factor of
for a 100 m thick membrane. The most penetrating size for a re-
cently tested sintered 316L stainless steel semiconductor filter was 0.1 m
with a penetration factor estimated at for a 1.7 mm thickness.
Metal filters, and to some extent ceramic filters, can simply be heated
from the outside with a small purge gas flow to carry off outgassing. Cau-
tion should be taken to heat ceramic filters slowly, however, as the ceramic
filter medium can be cracked by thermal shock. These new filters can be
heated to quite high temperatures with a purge gas (baked out) to achieve
ultrahigh vacuum levels of contaminant gas and moisture retention. Even
without bakeout, the metal and ceramic filters show much less outgassing
than PTFE. However, their high cost means that they will, for the
moment, be used mainly in small units near the point of use where they
may well be baked out in order to be a good match in performance to a
high quality point-of-use purifier.
Low-turbulence components and slow-opening valves. There is an expec-
tation that components will become available with lower turbulence factors
than hitherto, for example via smoother internal flow paths. These will
cause less entrainment of particles in cleanroom gases.
Putting restrictions or small orifices in pneumatic control lines to slow
the rate of opening or closing of valves has been shown to reduce the
number of particulates added to gas lines by valves. Some electric and
pneumatic actuated valves can be fitted with glycol-filled dampers to in-
crease opening and closing times. Replacing single valves with systems
comprising two valves and a laminar flow restriction element would also
seem to be a promising technique for achieving a slow opening characteristic
effectively.
Ultralow dead volume and integrated valves. Dead volume must be
avoided in ultraclean systems because it can accumulate particulates, make
purging by gas flow impossible and pressure-cycle purging slower. It can be
reduced by careful valve design and by incorporating the valves in a gas
panel design that minimises connection dead volumes. A major contribution
to the latter can be made by integrating several valves into one metal body
with several control orifices connected by small-bored passages. Such
integrated valves are now becoming more common on gas panels, despite
the limitations that they place on design flexibility.
Flow restrictors in pyrophoric and toxic gases. Flow restrictors reduce the
impact of a cylinder gas accidental release, allowing the accident to be
localised. Simple orifices are frequently used for this purpose. The only
alternative here is an excess flow valve of some description but this may not
be as intrinsically safe as a simple orifice flow restrictor. Flow restrictors do
lead to particles being released and injected into the gas used, however.
Measurements by USA gas company Matheson have put the contribution
(with inert gas) from a typical cylinder gas restrictor at c. 250 particles >0.1
m per cubic foot.
Reactive purifiers. These specialist purifiers are now being sold in larger
numbers, especially to the semiconductor industry. They mostly rely on a
microporous solid coated with a reactive chemical and are effective
absorbers of ppb water or oxygen. They do have problems, however. With
early models, accidental exposure to air would always result in almost
instant loss of capacity and could result in a disastrous overheating as the
reactive coating reacted exothermically with air.
More work needs to be done to test the long-term effectiveness of these
purifiers in different situations. For example, experiments with some early
models showed that for every molecule of water absorbed around 0.01
molecule of butane or other hydrocarbon is emitted from the purifier
material during impurity challenge tests. Some of the emitted hydrocarbon
may be bonded/dissolved further downstream within the resin of a purifier
unit but temperature cycling or simply a long time period could result in it
being emitted to the process. Particles of reactive material may be released
into the process gas stream, too, but these are easily stopped in the filters
normally downstream of the purifiers.
Claims are made for the latest generation of reactive purifiers (e.g.
Millipore Corporation Reactive MicroMatrix) that would seem to make
t hem the perfect device lor very small volumes of gas (they are very
expensive). For example, they wi l l not undergo melt-down if exposed to
air. The purifier formulations designed for need no pretreat-
ment or heating elements. This is an advantage compared to getters, which
need an initial heating cycle to activate or require continuous heating. They
are based on a metal carrier linked to a reactive metal such as lithium.
Purifiers are now available to purify some of the reactive gases such as
They are based on a
large surface area inert support coated with an organometallic reactive
compound. The organometallic is chosen to provide a suitable coating
capability and to react with the unwanted impurities water and oxygen to
produce an involatile oxide or hydroxide of the metal. Note that these
purifiers will not remove hydrocarbons. The purifier is first activated or
conditioned by flowing the gas to be purified through it. A typical example
is lithium triphenylmethane. This purifies, arsine for example, as
follows
Note that a small additional amount of is produced in the purifying
reaction, adding to the flowing in. In a similar way HC1 can be purified
with anhydrous support, inerts and hydrogen can be purified
with while can be used for fluoride gases such as
In the case of other hydrides, the pretreatment coats the purifier active
sites with bonded gas, e.g. groups for which are then released by
displacement by the preferentially absorbed impurities forming, e.g.,
again and absorbing
Purifier end-point detection. Several point-of-use purifier manufacturers
are now considering whether they can offer a simple analyser to give a signal
indicating the exhaustion of the purifier material. Currently, only large
purifiers include analysis of output purity. POU purifiers are simply replaced
on a schedule, normally ensuring they are only fractionally consumed. This
is both wasteful of purifiers and risky; an unforeseen air contamination can
easily result in premature exhaustion of the purifier. One concept is a very
simple quartz microbalance assembly, not based on reversible water
absorption as with the DuPont water analysers described in section 2.3.7 but
a cumulative instrument based on barium-coated quartz crystals; more than
a few ppb for a few minutes is capable of producing a frequency shift
measurable with low-cost electronics.
Oxygen passivation. Although it now seems that it really does work, it is
unclear that oxygen passivation is necessary for the current generation of
chip production. Oxygen passivation has been much more extensively
reported on recently and is offered by a number of, though not all, gas
equipment suppliers. The oxide film is formed by heating tubing or
components in a furnace at a few hundred C, while passing argon-diluted
oxygen through the tube or inside the components. The result after a few
hours is an oxide film which retains little moisture and other contaminants.
Although the oxide film is very thin and not particularly tough, it does seem
to resist the addition of monolayers of other molecules such as moisture.
Oxygen passivated tubing has been shown to dry down much more quickly
than standard electropolished tubing. On purging, moisture levels in
passivated tubing fall quickly to tens of ppt (1 ppt = 1 part in levels that
can only be measured with sophisticated APIMS techniques.
Ultraclean cylinders and aluminium. Aluminium continues to be the best
material for moisture performance. It is interesting to see that despite heroic
efforts with compound electropolished and passivated stainless steel
cylinders the ultraclean research team at Tohoku University have found the
same to date. Their heavy and expensive stainless steel cylinders cannot
really outperform a relatively ordinary high-quality aluminium cylinder.
Aluminium cylinders are typically specially treated, often by proprietary
anodising methods, which give them an unusually stable, pin-hole free,
unreactive and hydrophobic surface.
A consortium in Japan has been promoting and researching the use of
cleaned/passivated aluminium for vacuum chambers and pipework, evaluat-
ing aluminium components against stainless steel. There are now some test
results with hardened aluminium VCR-style joints for Al pipe available for
use with softened aluminium gaskets. There are, however, still problems to
be solved. It is possible that future ultrapure systems, as well as vacuum
systems, could use aluminium.
Cylinder coatings. Coatings have already been mentioned (sections 2.2.12
and 2.4.7) in connection with low-ppm gas mixtures and cylinder manufac-
ture. Semiconductor gases are always packaged in cylinders which have a
controlled inner surface preparation. Aluminium cylinders are anodised
with proprietary techniques whilst steel cylinders may also receive a
proprietary coating, often based on nickel. Chemical or electropolishing can
also be employed along with coatings.
Bags made from Teflon or a similar unreactive polymer could be installed
in cylinders to reduce contamination. Such bags would not save the cylinder
from corrosion necessarily (their permeability would be too high) but
particulates might be effectively reduced by this kind of measure. Although
not current technology, cylinders based on this concept were tried a few
years ago but without commercial success.
Continuous flow. Simply by sitting inside a gas line, for a variety of reasons
from dead-legs to outgassing, reacton or back-diffusion, gas becomes less
pure over time. When gas begins to flow from a stopped line into a process it
will not become pure for a considerable period. This is not satisfactory and is
one of the factors making the first batch of wafers different from subsequent
batches, which is always a headache in wafer fab plants. The best purity gas
lines are those which are in continuous use, the process gas itself purging out
any residual trace impurities.
Another argument for continuous flow is that cross-contamination can
occur because of back-diffusion when lines are inadvertently left open to
atmosphere at their downstream ends. The latter is often the cause of the
contaminants which are sometimes found on the downstream side of filters
in semiconductor process equipment. When the equipment is open to the
air, in order to admit a new batch of wafers, for example, or for routine
maintenance, the gas lines are switched off. Reactive materials such as silane
or chlorosilanes then have the opportunity to see small amounts of and
moisture with which they form particles. When the machine is restarted the
particles are carried into the process chamber. If the lines are switched from
a reactive to a purge gas, instead of switched off, this can be avoided.
3.5.6 Semiconductor deposition processes
Ultrapure single crystals of semiconductor. Before semiconductor device
fabrication can begin, the substrate, ultrapure monocrystalline silicon, must
be made. The bulk polycrystalline silicon from which IC grade single crystals
are grown is produced at a very few sites by companies such as Huels
(Monsanto, Dynamit Nobel), Union Carbide, Wacker, Shin-Etsu, Osaka
Titanium and Dow Corning.
The process starts by reacting HC1 with crude metallurgical silicon to
produce a mixture of chlorosilanes. Distillation of these produces the
feedstock for the silane plants and also the starting material for the
pure grades of chlorosilanes used in the wafer fab. For bulk silicon, these
chlorosilanes are then reacted with hydrogen. The Siemens process involves
heating a polysilicon rod, held in graphite connector chucks, to well above
950C to ensure reaction, depending on the silane or silane mixture used.
The starter rod, in the form of a U-shape in the classic process, is soon coated
with a thick layer of ultrapure polysilicon, which can be easily be broken into
chunks for crystal growing. The volumes of gases consumed are consider-
able but these processes are carried out at only a very few sites worldwide.
Impressively pure polysilicon from the above process is melted and then
grown into huge monolithic single crystals in an inerted furnace. The
Czochralski process pulls the single crystal out of the melt and effectively
carries out zone-refining during the growth. The growing of silicon crystals
takes place at very high temperature (the melting point of silicon is 1400C)
and a high quality inert atmosphere is essential. Argon is the gas of choice,
with helium being chosen occasionally.
Other semiconductors also need to be grown into single crystals. These
include gallium arsenide, gallium phosphide, mercury cadmium telluride,
indium phosphide and many alloys of these. The rule is that the ensemble
atoms in the crystal structure should have an average valence of exactly four.
Gallium nitride has recently attracted more widespread interest in optical
devices, whilst silicon carbide, until a few years ago only made for use as an
industrial abrasive, is showing promise for circuits, such as those used in car
engines, which must withstand high ambient temperatures. Most recently,
the element used in the first generation of semiconductor transistors,
germanium, looks like it may be able to stage a comeback in compound
layers with silicon. SiGe analogue circuits were recently demonstrated by
the IBM company and showed exceedingly good high-frequency perform-
ance. The use of a mainly silicon silicongermanium alloy means that circuits
can be processed in a similar way to silicon and fabrication is much easier
than for compound semiconductors such as gallium arsenide.
There are greater difficulties in growing compound semiconductor
crystals because of the problem of one of the elements escaping and spoiling
the necessary average valence. In gallium arsenide, for example, the arsenic
can sublime at 613C, whilst gallium, although melting at 30C, does not boil
until 2000C so heating up melts of GaAs tends to result in loss of arsenic.
Sometimes this problem can be avoided by crystal growth inside an
encapsulation of molten glass or by the use of high pressure gaseous
blanketing.
Epitaxial semiconductor layers. An epitaxial layer is a continuation of the
native single crystal wafer substrate into a deposited layer, typically of
higher purity than the substrate. Typical gases used are silane or chlorosi-
lanes for silicon and mixtures of organo-gallium volatile compounds with
arsine for gallium arsenide. Generally these gases are applied by being
flowed over wafers in a high-temperature furnace, dilution of the active
gases being employed. Silane is used for the lowest temperature deposition
processes (around 500C or so) with chlorosilanes requiring higher tempera-
tures, rising with their chlorine content up to 1250C for
Exceptional purity is required in silane gas for epitaxial deposition. The
epitaxial layer has to have a low intrinsic conductivity so that different parts
of the integrated circuit are not short-circuited. However, this low
conductivity can only be achieved if doping levels are extraordinarily low
and this requires ultrapure gas. Often silane gas is specified in terms of the
resistivity of the epitaxial silicon it is capable of forming. This is more
sensitive (ppt) than any gas analysis method.
The diluting carrier gas may be hydrogen (for chlorosilanes), occasionally
an inert gas such as argon or sometimes a ternary mixture. Vapours carrying
Group III or Group V elements such as boron and arsenic, respectively, are
added to achieve phosphorus or nitrogen doping in the epitaxial layer.
Reactors for epitaxial silicon, frequently used in advanced bipolar circuits,
are formidably large consumers of hydrogen and (chloro)silane. A wide
range of pressures can be used, from a few millibars up to atmospheric
pressure.
Today, single-wafer thermal reactors (RTP reactors, see below) are
available for epitaxial deposition, some with rotation of the wafer to assure
greater uniformity of the deposited epitaxial layer.
CVD reactor processes. Epitaxial growth from a silicon-containing vapour
is just one example of a chemical vapour deposition (CVD) process. A CVD
process is in general a gas-phase deposition of any compound onto a wafer
surface. CVD processes are used for depositing polysilicon (poly) used for
interconnection for circuits and for silicon nitride insulator layers. For
silicon nitride, mixtures with ammonia are employed. Although much
silicon dioxide produced is grown on native silicon (below), some stages of
processes require deposited oxide films, often created mixtures of silane,
nitrous oxide, oxygen and dilutant. CVD performed at a few millibars of
pressure is called low pressure CVD (LPCVD). These processes are more
modern in origin than the atmospheric pressure CVD processes and involve
the addition of large vacuum pumping systems to the reactor vessel. Many
CVD processes cannot guarantee good adhesion of the deposited film unless
they include a mild etch as an initial step. The mild etch step is often a brief
exposure to HCl gas.
CVD reactors can be either cold wall or hot wall. The latter type are
simply a type of vacuum furnace with a controlled atmosphere. They have
the snag that the CVD reaction happening on the wafer is happening on the
walls of the reactor, too, so that the latter will need frequent cleaning.
Cold-wall reactors employ RF inductive or radiant lamp heaters to heat the
silicon wafer and its carrier far hotter than the walls of the reactor, which
may be water-cooled to keep them reasonably near room temperature.
Deposition of the CVD compound on the walls is thus largely eliminated in
cold-wall reactors.
Common CVD reactions are those to deposit silicon dioxide (insulator) or
polysilicon (transistor gate) layers. Polysilicon is deposited using silane or
possibly chlorosilane although the latter does need a higher
temperature to deposit, typically mixed with hydrogen. The addition of
diborane, arsine or phosphine gas to the silane allows the production of
heavily doped polysilicon, which has a high electrical conductivity and is
used for polysilicon gate conductors.
Silicon dioxide-based insulator layers are typically laid down by mixing
silane with small amounts of oxygen near the heated wafer surface. This sort
of process can be carried out at pressures from high vacuum (a few millibars)
right up to atmospheric pressure. CVD glassy layers are often reflowed or
planarised, i.e. melted into compact layers with a smooth top surface, by
heating in an Ar annealing furnace if necessary.
Mixtures such as silane and oxygen, or silane and nitrous oxide, can be
doped with diborane or phosphine to deposit silicon dioxide based glass
dielectric material. If both are added, the dielectric is described as BPSG
(borophosphosilicate glass). These dopants make a glassy layer which can be
reflowed at low temperature. Glassy dielectrics are often used as insulator
layers, especially between metallisation, and are also used, laid down very
thick, in the final encapsulation (passivation or glassivation) layer that
protects the finished integrated circuit from environmental contamination.
These thicker layers are often put down in an atmospheric pressure system.
The pyrophoric nature of the gases used leads to the term pyro reactor for
some of the CVD systems used for this.
Vaporised liquid sources are now popular for dielectric CVD. Com-
pounds such as tetraethylorthosilicate (TEOS) are mixed with an inert
stream such as He or nitrogen in a heated bubbler. The TEOS stream
meets an ozone stream near the wafer surface and silicon dioxide is
deposited on the silicon wafer, which is heated, typically to 400C. A 500 nm
layer can be grown in 5 mi n. TEOS is much less dangerous to handle than
silane, being only a moderately reactive liquid. Dopants to make the glass
reflow at low temperature are vaporisable liquid sources of phosphorus and
boron such as trimethyl phosphite and trimethylborate.
More and more exotic materials are being used in CVD in order to utilise
the fast deposition, high-purity characteristics possible with good CVD.
Starting with the less, and working toward the more exotic, some of the
more well-known compounds are:
tungsten hexafluoride for tungsten conductor via plugs forming

connections between different conductor layers in the integrated circuit;
tetrakis(diethylamido)titanium/ammonia: for t i t ani um nitride barrier

layers as an alternative to nitriding a sputtered titanium layer;
tripropyl aluminium: for Al conductors, instead of sputtered Al films;
various copper metallorganics (e.g. copper(II)beta-diketonates using

argon or hydrogen as a carrier gas for copper conductors): replace Al
conductors in future chips needing the ultimate electrical conductivity of
copper;
tantalum pentaethoxide: for producing films of tantalum pentoxide, a

high dielectric constant material, useful for increasing the data storage
capacitor values in DRAM (dynamic random access memory) circuits.
Most of these compounds are vaporisable liquids, not gases. They are
delivered to the wafer surface by their carrier gas, often inert (argon or, for
lower temperatures, nitrogen) but often, too, a reducing carrier (hydrogen).
The carrier gas is bubbled through the liquid compound with some
precautions to avoid too much liquid droplet carry-over and temperature
control of the bubbler to ensure a constant partial vapour pressure of active
compound. Further carrier gas is then mixed in downstream until the desired
concentration is attained. This additional dilution also avoids condensation
problems.
Some of the compounds need to be in relatively high concentration in the
carrier gas and this can lead to problems with liquid condensing out in the
distribution pipeline. In this case, the gas lines (and supply cylinder, too,
perhaps) will need to be heated, with a positive temperature gradient
towards the point of use. Low-temperature trace heating by heating tapes is
a suitable technique, if combined with suitable lagging so that cold spots are
not formed. (It should also be noted that chlorosilanes can also condense in
cool tubing under pressure and similar precautions are occasionally
necessary even with gases.) Alternatively, the supply bubbler can be
cooled or the gas can be delivered at subatmospheric pressure. Special
regulators are made with large orifices which will work down to a few
millibars of absolute pressure.
Selective CVD. CVD reactions are typically blanket reactions, that is,
they form a layer, or blanket, right across the whole wafer. This does not
have to be the case and reactions such as tungsten deposition can be
operated so that they take place selectively; for example, depositing on Al or
silicon areas on the wafer but not on areas. Sometimes reacting the
wafer surface in some way, by etching it mildly, for example, can render an
otherwise non-selective process selective. These selective processes are
more difficult to operate, however, and are mostly slower than the blanket
reactions. Thus tungsten deposition for via plugs is most often achieved
today by blanket tungsten deposition followed by etch back until only the
plugs are left.
Diffusion doping of semiconductors. Semiconductors in the pure state are
almost insulators. To make them conduct electricity, they must be doped
with small amounts of elements which replace some of the silicon atoms but
also donate electrons or remove electrons from the crystalline structure. The
donated electrons render the semiconductor conductive with negatively
charged carriers (n-type). Less obviously, where electrons are removed the
other electrons in the crystal behave in such a way that they simulate mobile
positive charges, called holes, in p-type semiconductors. A typical p-type
doping element is boron, the atoms of which have three electrons in their
outermost electron orbitals. Boron replaces a silicon atom while removing
an electron. A typical n-type doping element is phosphorus, which donates
an extra electron to the crystal, having five electrons instead of four in the
outermost orbital.
Doping gases are thus those containing elements from the III and V
columns of the Periodic Table for silicon, and the II and VI columns for
gallium arsenide. Phosphine and diborane are typical, being supplied mostly
in dilute mixtures with hydrogen. Only tiny amounts of the elements are
required as the doping concentrations are in the parts per billion range.
Purity of the gas mixture is therefore of strong importance, otherwise
inadvertent doping or undesirable dopants can be introduced. The doping
gases are often used in a diffusion furnace, in which atoms of dopant are
deposited on the semiconductor surface and then driven into it by being
heated so that solid state diffusion takes place. Dopant atoms can also be
incorporated by addition of doping gases to the silane-based gas mixture
used to deposit polycrystalline or epitaxial silicon.
Ion implantation doping of semiconductors. A more modern alternative
technique for doping, which uses very tiny quantities of dopant, not always
as gas, is ion implantation. Here beams of dopant ions are created by a
plasma ion source and then accelerated into a high-vacuum chamber by
electrostatic fields. The beam of ions, which may have energy anywhere in
the range of a few tens of keV to 2 MeV or more, is then impacted into the
target silicon wafer. An electron beam gun ensures that the wafer surface
does not become charged up. The plasma ion source is typically fed by tiny
flow rates of hydride gases such as arsine, diborane or phosphine. The ion
beam currents used commercially can be up to 100 mA or more, with limits
on beam current depending on ion repulsion defocusing effects at low ion
energy and on heating effects at high ion energy.
In principle, the purities needed are not high since ion implanters
normally transport the ion source output via a magnetic focusing and
separation beamline to the target. The beam transport is in fact a mass
spectrometer, allowing only ions of the correct charge/mass ratio through
onto the wafer surface.
Oxide growth. When a layer of silicon oxide is required on top of silicon, it
is produced by oxidation of the wafer silicon or by addition of silicon
dioxide. The former process is preferred for high quality oxide. There are a
number of variations on the silicon oxidation reaction used in the industry:
wet oxidation: uses steam + oxygen
dry oxidation: uses dry oxygen
HIPOX (high pressure oxidation): uses oxygen at high pressure (although

this process is currently largely obsolete).
HCl is often added to accelerate the rate of oxidation and improve the oxide
quality. These processes are all fairly modest users of oxygen, a small LO
tank supply being the norm. One variant of the wet oxidation process is a
more substantial user of gases, however. This is the variant pyrogenic
steam in which hydrogen is burned with excess oxygen to provide high
purity steam plus oxygen for the furnace oxidation atmosphere. The
consumption of oxygen and hydrogen can be considerable.
CVD-deposited oxide can be formed at temperatures as low as 200C. The
grown oxide process and its resultant silicon dioxide film is often referred to
as thermal oxide because it is formed only at very high temperatures (above
800C). The wafer is heated to an accurately known temperature in the
oxidising gas stream for a precise period of time and then cooled until further
reaction is not possible. Oxidation reactions can take many hours or just a
minute or two because the speed of oxidation is a sensitive function of
temperature and oxide growth is not linear. Thick oxides are particularly
slow because oxygen must diffuse through a thick intervening layer of oxide
to reach silicon to oxidise towards the end of the process. The oxidation
reaction is so important to silicon semiconductors that it has been
extensively studied.
The reaction is governed by two thermally activated processes:
1. diffusion of oxygen atoms through the already-formed silicon dioxide;
2. reaction of oxygen atoms at the silicon/silicon dioxide interface.
If is the activation for the first process and is the activation for the
second, the oxidation thickness X takes the following course versus time, t
Where D is the diffusion constant for O in Si, (where
is a constant), K is the reaction rate constant for oxidation
the constant is 5 or 10 times greater with or HCl than
for dry oxidation), is the surface concentration of (related to partial
pressure), is a constant (O atoms in unit volume of and is a
constant related to initial oxide thickness.
For small diffusion times or low temperatures oxide growth is reaction-
rate limited and X follows a linear law
For longer diffusion times or higher temperatures oxide growth is diffusion-
rate limited and X follows a square root law
(This kind of model also applies to many other solidgas reactions, such as
the formation of oxide on metals in air. Metals which oxidise slowly in air
generally form stable, continuous oxide layers, e.g. silicon, which have a low
diffusion constant. These metals, such as aluminium or stainless steel, reach
the diffusion-limited regime in seconds and further oxide growth is almost
imperceptible. Metals which oxidise rapidly are generally those which do
not form such a diffusion barrier. The oxide formed is lost by flaking or some
other process and they oxidise continuously in the reaction-rate limited
regime. The picture is more complex than with silicon because of catalytic
effects, electrochemical cells and other corrosion reactions in the presence
of moisture and other atmospheric contaminants such as
A typical industrial process in a conventional quartz tube diffusion
furnace might operate at 1100C for two or three hours to obtain 1000 nm
field oxide, using wet oxidation. With single wafer rapid thermal processor
dry oxidation, a much shorter time in a hotter furnace is used, e.g. 1200C
for 200 s for 40 nm of gate oxide.
The purity of oxygen used in these processes, formerly not regarded as
important, has recently become an issue. One of the problems is that even
cryogenic oxygen typically has 2040 ppm of hydrocarbon impurities, which
can lead to harmful carbon in the oxide; carbon typically lowers oxide
breakdown voltage, important in the ultrathin gate oxide used today.
Oxygen purifiers are now used in new wafer fabs.
Rapid thermal processing. One technique that is slowly growing is that of
using high speed single wafer furnaces, i.e. rapid thermal processing (RTP).
Here a single wafer is loaded into a (typically) quartz furnace, purged for
30s, heated in approximately 30 s up to operational temperatures (up to
1100C), and then cooled in a flow of inert gas back to 200C, at which
temperature handling is possible, in a further 30 or 60 s or less (see Figure
3.24). This rapidity of processing is possible only by using banks of
tungsten/halogen lamps for heating, with pyrometric temperature measure-
ments providing feedback with which to control the heating rate, evenness
of heating and final temperature. The speed at which processing can be
carried out, with only 2 min or so per wafer, means that RTP can be
competitive with multiwafer furnaces taking 25 or 50 wafers but needing
hours to process.
RTP can potentially be used for many processes normally carried out in
large multiwafer vertical or horizontal tube furnaces, e.g. silicon epitaxy,
gate oxidation, CVD silicon dioxide, annealing or CVD tungsten. However,
only a small number of operations routinely use RTP today. An example is
titanium silicide layers, formed by reacting sputtered titanium with wafer
silicon at 700C, where the controllability of the conditions in a single wafer
reactor have allowed it to take over from more traditional furnace
treatment.
There are a number of problems with RTP; some fundamental, some
more in the nature of teething troubles for a technique that is still not quite a
standard production technique. In the first category is temperature unifor-
mity, especially during ramp-up and ramp-down. Evenness of lamp
illumination is never quite perfect. Different parts of a part-processed
(patterned) wafer absorb IR energy from the lamps at a different rate and
heat up differentially, an effect so pronounced that some wafers will curl up
visibly during the heating cycle. However, the latest RTP designs claim to
Figure 3.24 RTP cycle.
have solved this problem, getting around 2C uniformity across a wafer.
Another furnace avoids the problem of evenness of lamp-heating by simply
pushing the wafer into a constantly hot furnace section at the top of a vertical
tube furnace, achieving wafer processing with a cycle time only a few times
slower than tungsten/halogen RTP.
Another fundamental problem is that of purging of moisture. The
constraints of the short cycle time are such that it is difficult to purge out
moisture completely from the system down to the ppb levels possible in
larger multiwafer systems. As noted above, wafers themselves have water
absorbed onto their surface when they are first put into the RTP at room
temperature.
Sputtering for metallisation. The sputtering process basically involves the
bombarding of a metal target surface with argon ions, releasing metal atoms
which can then be deposited in a layer on the IC surface, the whole process
taking place at lower pressure than CVD Sputtering is
popular because it yields higher quality films than other processes, such as
vacuum evaporation. It is a user of argon gas, which is the favoured gas for
generating a stable glow discharge.
As well as straightforwardly striking a glow discharge between the metal
target and the wafer, it is possible to extract positive ions from the argon to
direct at the target, an arrangement often used at lower pressures, higher
powers and for dielectrics. The glow discharge can be made very much more
powerful by using RF, with 10 kW and more power supplies being in use to
speed up the process. The wafer being coated is often heated to a few
hundred C. This gives a better film, presumably because it allows some
thermal relaxation of the initial landing position of the sputtered atoms at
lower temperatures the atoms freeze where they land. However, IC
manufacturers try to minimise the amount of thermal exposure that a wafer
is subject to late in its processing, so sputtering at 200300C or less is now
normal.
Sputtering is a relatively slow deposition process, although speeds have
increased dramatically with the advent of high power RF plasmas, and is
mainly used for metal films (Al and Ti are typical requirements). The
sputtering process, being largely a physical phenomenon, also works on
metal alloys and on dielectrics by using a target made of these and adjusting
process parameters. Typical sputtered thicknesses are low, e.g. 20500 nm,
with deposition rates of the order of in powerful machines being
typical of the maximum rates attainable on a metal like Ti; some materials
are, however, much slower to sputter.
Problems with sputtering are few. A high quality vacuum is necessary as
otherwise the vacuum residual gas will be incorporated in the sputtered film.
High-purity argon is indicated, backfilled into a high achieved quality of
vacuum. Sputtered material covers the entire inside of the reactor vessel,
not just the wafer being processed, so the machine must be regularly
dismantled and the deposits removed, normally by suitable chemical
processing, or replacements substituted for the coated parts. By careful
design, inexpensive shield parts can intercept nearly all the surplus coating.
These shield parts are quickly replaced, being either disposable or easily
cleaned on a wet bench. Flaking off of the vessel coating is a major potential
source of particulate contaminants. Some sputtered films on the reactor
vessel are so firmly attached that removal is not necessary. Other films are all
right in place provided they are not oxidised. Small amounts of oxygen
impurity in the gas can lead to oxidation and flaking of the oxide formed; the
use of ultrapure gases and clean vacuum can avoid this problem.
Reactive sputtering is sputtering carried out in an atmosphere which can
react with the newly sputtered film. Titanium nitride, for example, can be
laid down by sputtering titanium on to a wafer in an atmosphere of argon
with a few tens of percent nitrogen at mbar pressure.
3.5.7 Plasma etching: a key technology
The art and science of plasma etching has been one of the most important
innovations in the semiconductor industry. Plasma etching (sometimes
called reactive ion etching or loosely dubbed dry etching) is simply a
means of removing a thin layer material from a wafer surface. It can be
non-selective, that is, removing the surface whatever the material there, or it
can be selective, removing metal while leaving insulator or vice versa.
Selectivity can be defined as the rate of desired etching divided by the rate of
unwanted etching, minus one (although the one is often not subtracted).
Etching can also be either isotropic or anisotropic, terms which warrant
some explanation.
The etching of patterns on a silicon wafer in general is achieved by using
an etch process which removes substrate but not the photolithographically
patterned polymer resist material. An isotropic etch simply etches the
Figure 3.25 Isotropic (top) and anisotropic (bottom) etching.
substrate (but not the resist) in all directions at the same rate. This means
that the resist will be undercut about sideways for every of
downward etch. Anisotropy can be defined as the rate of vertical etch
divided by the rate of sideways etch, minus one, though again the latter is
often omitted. Anisotropic etching occurs where several effects, such as
plasma bombardment and the lifetime of excited species are significant and
areas which can hide behind the polymer resist will not be etched much, if
at all, so that, for example, a etch downwards might only undercut the
resist (if at all) by a small fraction of a micron (Figure 3.25). Anisotropic
etching has big advantages over isotropic etching, particularly with the small
size of circuitry now being sought by integrated circuit manufacturers.
Anisotropy allows structures such as deep but narrow conductors and allows
much denser packing of circuits. High anisotropy is one the key achieve-
ments of modern plasma etching processes.
Isotropic etching. Simple dry etching in which a plasma is not used, using
an anhydrous reactant gas, cannot possess the advantages of anisotropy but
can be cleaner and more controllable for small etch rates than wet etching.
Etchants include HCl, HF and chlorine. These processes are sometimes
called thermal etching, as it is usually necessary to raise temperatures. The
term dry etching is often used more loosely to refer to plasma etching,
which is also a dry process.
Dry-etching processes, such as plasma etching, are popular anyway
because they avoid the handling and purity problems of wet chemicals.
Plasma processes are even more popular because they open up the
possibility of anisotropic etching, which is increasingly important with the
shrinking geometries now evolving. The use of an electrical discharge to
create excited atoms or ions is the basis of the plasma etch technique, the
excited species being derived from gas mixtures (Table 3.8).
Dry-etch gases include fluorine and chlorine compounds such as the
CFCs, chlorine, mixtures of these with oxygen, sulphur hexafluoride and
more exotic materials such as silicon tetrafluoride or boron trichloride.
Reactive ion etching (RIE) employs lower pressures at which these gases are
more substantially ionised and at which more anisotropic etches can be
obtained. Argon is frequently employed as a chemically inert diluent gas,
and, as usual, as a gas in which plasma formation can be easily obtained.
Hydrogen may be added in order to help form volatile hydrides of etch
products.
The simplest plasma reactors place a silicon wafer on a platform consisting
of one electrode facing a flat-plate electrode. Plasma conditions, a highly
ionised gas, are produced in plasma reactors by using low pressures (from
0.1 mbar or less to tens of mbars) and an RF (e.g. 13.56 MHz) or microwave
(e.g. 2.5 GHz) alternating electric field. Efficiency of plasma production is
raised by carefully tuning the RF circuits or, in the case of microwaves, by
adding a d.c. magnetic field to give an electron cyclotron resonance
enhancement. The electron cyclotron resonance (ECR) mechanism deflects
the electrons by using a magnetic field such that one (or more) orbit of an
electron in the magnetic field takes place at about the same frequency as the
applied electric field. The cyclotron effect occurs because the orbital time of
the electrons is constant, no matter what their energy, because as their
velocity increases, the radius of curvature of their orbit increases pro-
portionately (until relativistic velocities are reached).
Etchers can be designed to etch boats of 25 wafers, but it is difficult to
ensure etch uniformity. Normal designs either etch one wafer at a time
against a parallel planar counter-electrode or etch up to a dozen or so
clamped on a rotating polygonal prism (hexode) or on a dodecahedron
with stationary counter-electrodes around.
The electric fields accelerate natural ions and an avalanche of (mainly
positive) ions and electrons quickly forms an equilibrium state where 1 in
of the gas molecules are ionised. The ions and electrons are
fast-moving, much more so than the gas temperature would suggest, and
this is the key to part of their activity. They collide with the other gas
molecules, breaking apart the molecules and releasing free radicals and
free atoms as well as further ions. A space charge of ions builds up which is
positive, repelling ions from itself so that an RF plasma tends to have ions
flowing away from its middle.
Anisotropy, selectivity and etch rate are the key parameters of plasma
etch processes. Selectivity of 100:1 of polysilicon on is possible, whilst
etching at just 4 to the vertical is possible for sidewalls (corresponding to a
high degree of anisotropy). Etch rates vary enormously, depending on the
layers being processed on the wafer. Suffice to say that plasma-etched
trenches of 50 or are possible for special applications such as
microengineered devices, although normal etch rates might be only
per minute or less. In addition to these primary parameters, the effects on
photoresist coating, the deposition of particles or polymer (the latter
derived from the materials being etched and the photoresist), the surface
damage and the surface smoothness of surfaces after etch are also
important.
The ideal plasma is one which has a very high density of low energy ions
while operating at low gas pressure. The low energy of the ions will avoid
disrupting the atomic structure of the semiconductor surface whilst the
high density assures a high etch rate and anisotropy. However, designers
strive to achieve this using different designs or plasma generation. They
have the problematic compromise that high power, needed for high
density, tends to produce high energy ions, causing damage unless gas
pressure is increased, when anisotropy decreases. Etch residues are also a
concern since, if the products of the etch reaction are not volatile and
speedily removed, there will be problems.
The process variables which may be manipulated are the gas com-
position and pressure, the RF field, the d.c. voltage on the semiconductor
wafer, the wafer temperature and the reactor geometry. Plasma processes
are, in industrial practice, largely empirically determined simply because
the science is complex. The effect on anisotropy, selectivity and etch rates
of a change of process variables cannot be predicted in general. Plasma
etching is of such industrial importance that a huge amount of effort has
been put into producing workable processes. Processes and equipment are
now available to carry out almost every conceivable etch job, although with
some compromises on performance, despite a largely empirical basis.
Much work over the years has extracted the underlying mechanisms,
however, and these are used as a guide to predict the process parameters at
least approximately. Plasma etching uses several basic physical effects as
follows.
Sputtering: the mechanical dislodging of surface atoms by impinging high

speed ions and neutrals (section 3.5.6). Sputtering is highly directional, so
is important in anisotropic etches but is normally very slow. It fails to
account for the speed of plasma etching by many orders of magnitude.
Chemical etch by species such as free atoms and radicals produced in the
plasma. Straightforwardly, for example, F atoms produced in a plasma
can attach to silicon atoms and form which is volatile and lost from
the surface. Chemical etch is isotropic, at least for longer lifetime species,
so the chemical etch mechanism must be less important in highly
anisotropic etch processes. In downstream etching, such as stripping
processes, presumably chemical etch by excited species is the only
mechanism.
Collision-energy induced chemical reactions are reactions which occur

only because the bombarding molecule or atom is moving fast enough to
damage the etched wafer. The surface of the semiconductor wafer being
etched is excited by the collision in some way; it may have a dangling
bond, for example. The surface at that site will then react readily with the
gas present or with another bombarding ion. This mechanism leads to
high anisotropy; only surfaces bombarded are etched in this way, surfaces
hidden by photoresist are not excited and not etched.
Sidewall protection occurs when the natural chemical etch reaction

produces an involatile species which forms a protective layer on the wafer
being etched, stopping further etch occurring. The tough protective oxide
layers which protect stainless steel and aluminium from oxidation in the
air are familiar everyday examples of this effect. However, in the plasma
etch reactor the protective film can be removed by the action of
bombardment by the sputtering process or by ion-collision induced
chemical reactions. This mechanism does not affect the sidewalls, whose
protective film stays largely intact. Sidewall protection also leads to high
anisotropy. Examination of sidewall films sometimes reveals this mechan-
ism at work; for example, fluorocarbon-based polymers are sometimes
found after etches using
Choice of etch gases is determined partly by product volatility. For
example, for etching A1 interconnection layers, chlorine-containing
CFCs or are used because aluminium chloride is relatively volatile.
When etching blanket tungsten (used for interconnect vias between
layers), F-containing CFCs, PFCs or for example, are used because
is volatile. Oxygen is frequently added to mixtures because the basic
reaction is an oxidation where, for example, metals are being etched. Gases
such as chlorine are themselves oxidising agents and mixtures with chlorine
will not require oxygen addition.
Problems with plasma etching. Although plasma etching is now ubiqui-
tous, it has had to overcome many problems. A few are described below.
Figure 3.26 Plastic film strips being plasma-treated for surface modification (courtesy of
Oxford Plasma Technology).
Etch start is sometimes inhibited on silicon and metals by the presence of
native oxide film, i.e. the oxide formed while the wafer was sitting around
in the cleanroom air. This can be avoided by adding a minute amount of an
oxide-removing gas to mixtures; fluorine-containing molecules would work
for the native oxide which forms on silicon surfaces preventing silicon etch,
for example.
Particulate formation can be a big problem in plasma reactors. As
particulate contamination specifications for integrated circuits move ever
downwards, even small numbers of minute 0.2 or particles on the
processed wafer may be unacceptable. Reactions in the plasma may create
particulates, which is bad enough but might be acceptable if they could be
immediately swept away by the gas flow. However, it is frequently not so
simple as this. The plasma charges up all particulates by adding electrons to
them and the charges generated may then allow the particles to become
trapped in the plasma, and be lost only slowly by removal mechanisms such
as gas flow, deposition or evaporation.
First observed by astronauts near their spacecraft in orbit, particles
trapped by plasma can easily be seen in a low power plasma by, for example,
shining a laser beam through the chamber during operation and observing
with a narrow-band red filter. Video sequences were first reported by
Selwyns IBM group in the USA and showed this curious trapping
phenomenon starting up. Small particles accumulate, suspended in
patches characteristic of the reactor geometry and RF fields. Small clouds
are often found near changes in the reactor electrode surface, such as the
edge of the wafer. Often flakes or needles form, each made up of thousands
of particles agglomerated together. Deposition of trapped particles on the
wafer being processed may occur when the plasma is switched off and the
charged particles are dumped on the wafer either by gravity, migration or by
electrostatic forces.
Monitoring of particle densities is an important diagnostic in any vacuum
reactor. Equipment is available (e.g. the USA-based HYT Corporations in
situ particle monitor) which uses right angle scattering from a laser beam in
the exhaust line from a chamber. These particle monitors are rather like an
atmospheric OPC (optical particle counter). The detection chamber uses
mirrors to create a net of laser beams across a wide (50 or 100 mm)
aperture, however, rather than forcing the sampled gas stream through a
narrow constriction illuminated by the laser.
Excessive polymer deposits can often be caused by polymerisation of
fluorocarbons. This can be controlled by reducing the fluorocarbons, using
another fluorine source, or adding oxidant. High flow rates can carry away
polymer precursors and hence suppress deposits.
The ions bombarding the target, i.e. the wafer being processed, cause
charge build-up on that wafer. The resulting voltage on the wafer is called
bias and can affect the etch effect markedly. Ion implantation can also
occur, causing sub-surface damage to the wafer. At lower pressures the bias
voltage can be considerable, resulting in ions being accelerated into the
wafer leaving trails of ionisation and disruption of the silicon crystal below
the level being etched. This damage can be minimised by operating at a
lower bias voltage condition with plasma conditions that minimise the ion
energy. The latest designs achieve 20 eV ion energies, whilst maintaining ion
densities in the plasma of 10
11
or 10
12
near the wafer surface in a low pressure
(0.010.05 mbar) gas.
Next to particulate contamination, non-uniformity is probably the biggest
problem of many semiconductor processes and leads to many rejected ICs
from a wafer. An ideal process etches every wafer, and all of that wafers
surface, to the same degree. For processes like photoresist strip this may not
be critical in that the etch process will be largely stopped once the resist has
cleaned off and over-etch can be generous. With processes with a less
effective etch-stop, however, over-etch must be minimal and the uniformity
of the process must be as high as possible. Clearly, single-wafer diode and
hexode etchers will have important advantages over boat etchers here and
the boat equipment is only used for non-critical stripping processes.
Improvements to the flow patterns inside reactors can improve uniformity,
as can use of guard areas on flat-plate etchers and rotation of the wafer
holder. Temperature control of the wafer is important. Many etchers now
use a cooling system, such as flowing helium, on the wafer backside.
Uniformity is generally enhanced at lower plasma power levels, an effect at
least partly due to the more uniform temperatures achieved. Roughening is
also a kind of non-uniformity. Roughening of an etched surface is often seen
and needs to be controlled and limited if it is not to cause problems.
End-point detection. It is very important to be able to stop a plasma
process when its etching is finished. Commonly, the plasma discharge will
change its electrical characteristics near to the end-point. The effect of the
plasma becoming less conductive, for example, modifies the load seen by the
RF power supply; sensing changes in the latter enables an end-point to be
declared. With increasingly difficult processes using plasma, and with less
selective processes being used more, however, such a simple approach is
now much less useful.
A more sophisticated, but expensive, approach, is to attach a small
quadrupole mass analyser to the plasma rig and look for the presence of
tell-tale ions which indicate the end of an etch, perhaps simply by seeing a
large increase in ions from the substrate at the etch bottom. However, the
addition of a mass spectrometer system requires a large intrusion into the
vacuum system in order to be able to sample the plasma effectively. The fact
that plasma emits a typically bluemauve glow, which changes slightly with
the material being etched provides an easier means of monitoring. Optical
emission from the plasma can be used as an end-point detection system,
which is more easily retrofittable. It is now common to monitor plasma via its
optical emission, sometimes using two emission frequencies: one a wave-
length indicating end-point, the other a wavelength providing normalisation
by giving a measure of the overall plasma power.
3.5.8 Other plasma processes
Resist dry stripping. Resist dry-stripping, often called ashing, is a
specialised kind of isotropic plasma etching that is in many cases replacing
wet solvent methods. The polymer-based photoresists used during silicon
processing must be removed after the main etching process. They are often
surprisingly difficult to remove. After coating, exposure and development
they have to be hard-baked on by heating strongly, perhaps also with UV
light, converting them into the very tough polymers needed to resist
powerful etch processes. Plasma etching may leave polymeric deposits
mixed with silicon, metals and minerals along the sidewalls, and sputtering
can also occur, further modifying the resist. Other processing, such as ion
implantation, can cause a toughened, even carbonised, skin on the resist
surface. Even with mixtures of powerful solvents such as N-methyl-
pyrrolidone with water and ethanolamines, and even with heated solvent
mixtures, there will often be residual resist adhering to the silicon wafer
after wet-stripping.
Dry-stripping is a common solution to these problems. Oxygen atoms
are generated and these combine vigorously with all the resist components
while etching the wanted layers only slowly. An oxygen-based low pressure
(120 mbar) plasma is employed, pumped by RF or microwave energy,
usually with a fluorine-containing gas such as or a CFC. As well as
processing systems with the plasma generated immediately above the sili-
con wafer, downstream plasma generation is possible.
Alternative schemes have employed ozone in oxygen at atmospheric
pressure or nitrous oxide as the oxidant in a low-pressure plasma system. A
variant of the ozone generator system normally used in water treatment
(section 2.2.7), silent electrical discharge, can be used. Illuminating the
wafer surface with UV light to further activate the ozone by breaking it
down into oxygen atoms in situ can also be carried out. Often, resist dry
stripping is followed by a relatively mild wet-chemical etch, sometimes just
a deionised water rinse, to remove any mineral and other non-organic de-
posits which the oxygen atoms have not removed.
Plasma-enhanced CVD (PECVD). PECVD uses the plasma excitation as
an energy source to drive deposition of materials from a gas mixture on to
the silicon wafer surface. A large number of traditionally CVD processes
can be moved to a PECVD process, often with the advantage that a less
aggressively reactive gas or vapour can be used. PECVD is a process which
is growing steadily in importance in the semiconductor industry.
One recent example of this is the use of PECVD for creating low dielec-
tric constant insulator layers made from fluorinated silicon dioxide (these
speed up circuit functions by reducing the electrical capacitance of inter-
connections). A silicon source, either tetraethoxysilane or silicon tetrafluo-
ride, with oxygen and, if necessary, a fluorine source such as nitrogen
trifluoride or perfluoropropane is used, at a few millibars pressure. In this
way, when the plasma is struck, layers of of insulator can be de-
posited. The insulator typically has 510% fluorine content and a dielectric
constant of 3.5 or so, giving a useful advantage over pure silicon dioxide
whose dielectric constant is typically 4 or more, depending on density and
doping.
Dry cleaning processes for silicon wafers. Many of the process steps that a
semiconductor wafer must pass through during fabrication are cleaning
steps and most of these are still performed with wet benches. Wet benches
are sets of baths of ultraclean chemicals and deionised water, sometimes
automated with the addition of a robot handler for boats of wafers. Broadly,
the cleaning steps carried out on wet benches amount to breaking the
electrostatic adhesion of particles with an aqueous liquid, oxidising the
impurities or etching back the wafer slightly, releasing surface particles and
removing native oxide (thin oxide formed on exposure to air). Popular
cleaning baths are named Piranha (a hot sulphuric acid oxidising process),
SC-1 (an ammonia akaline oxidising bath) and SC-2 (a hydrochloric acid
oxidising bath). Others are known by their chemistry, as in buffered HF
(HF plus ammonium fluoride) or HF dip.
Wet benches consume large amounts of chemicals even though recircu-
lation of the bath chemicals through filters reduces consumption. The use of
hydrogen peroxide as an oxidiser is inconvenient as is relatively
dangerous, expensive in high-purity grades and decomposes easily. Spray
cleaners, which are economical with chemicals, are still trailing wet benches
in performance for most cleaning applications and are only used for less
stringent cleaning requirements. As a final disadvantage, silicon wafers must
be dried after each wet-cleaning process. The drying process causes delays
and has the potential to concentrate any contaminants from the last rinse
bath onto the wafer surface.
Supplanting wet-cleaning steps might improve yields and would make the
wafer fab safer and less expensive to run. No generally applicable and
accepted process is available currently to replace wet cleaning but it is the
goal of a number of research projects worldwide to produce an alternative
dry-cleaning process, i.e. a process for cleaning wafers without using the
usual mixtures of unpleasant wet chemicals.
A few proposals simply use a reactive gas mixture. HF gas has been shown
to contribute a cleaning action and, like hydrofluoric acid, leaves the wafer
passivated with F atoms, preventing oxidation in the cleanroom air. The
problem of having to handle HF, however, is significant. Gas companies, as
well as vacuum equipment and semiconductor equipment companies, have
joined these research efforts, such is the importance of such a process. UV is
being tried as the source of excitation for oxygen mixtures, as well as plasma
sources, and many different combinations of gases are being researched.
Hydrogen and argon downstream plasma arrangements are being tried in
which the gas mixture passes through a plasma discharge and then over the
wafers, avoiding problems with ions from the plasma and allowing only
relatively long-lived excited species to reach the wafer surface. Both excited
Ar and atomic hydrogen are known metastable species with high reactivity.
A demonstration of the downstream plasma process is occasionally seen in
undergraduate chemistry laboratories. If an organometallic with a high
vapour pressure, such as lead tetraethyl, is pyrolysed at a section of a quartz
tube by heating strongly, a shiny metallic film is laid down. The pyrolysis
yields free radicals. The deposited film can be cleaned away by heating the
tube again but this time upstream, exposing it to a stream of vapour containing
free radicals from the decomposition.
Cleaning with liquid blast. One unusual idea that seems to be
inapplicable to silicon wafers but might well become popular in other
applications, is the use of a blast of gas/liquid/solid. Ultraclean liquid
under its vapour pressure of 50 barg or so, is directed at the surface to be
cleaned via a small carbide/teflon nozzle assembly. The resultant high-speed
stream of gas, containing fine snow particles and probably also
(fleetingly) liquid, is very effective at removing contamination from smooth
surfaces, even down to very small particles. The technique is in use in the
manufacture of photoconductive drums for photocopiers and laser printers,
and has been proposed for cleaning the inside of gas cylinders for ultrapure
gases.
Fingerprints and oily marks as well as particles are removed by
However, there are problems. The used must be of a special high purity;
standard is of a much lower purity than other cryogens and a special grade
is needed. Another problem is that the substrate being cleaned is cooled down
considerably. In air, water will be condensed onto the surface almost
immediately, presenting another potential contamination problem. Cleaning
of parts inside a dry box filled with dry air or nitrogen avoids this problem but
throws away some of the techniques convenience.
There is a further proposal to use under pressure as a supercritical fluid
for the development of photoresist. It turns out that certain substituted
polymeric silanes, such as poly(methyl phenyl silane), are depolymerised to
siloxane by UV light and this exposed resist material can be removed by a
solution of the siloxanes in supercritical
A dry-cleaning process that is now an accepted technique is the use of a glow
discharge or low-power plasma for cleaning the inside surfaces of vacuum
equipment. Depending on the contamination, mixtures containing oxygen
and fluorine or chlorine compounds are the most popular. For hydrocarbon
contamination, such as might be seen in a plasma etch system working on
photoresist, a mixture of with oxygen is common, although with etch
chambers that are contaminated with aluminium a chlorine-containing etch
mixture is required. A mixture of with chlorine can both oxidise
hydrocarbons and form volatile aluminium chloride, so solving this problem.
3.5.9 Semiconductor effluent gases
CFCs, PFCs and toxic emissions. In recent years processes have been
developed which largely avoid using CFC gases which can damage the
earths ozone layer (section 2.5.12). The Montreal Convention will reduce
CFC production and emission; to zero for some compounds and to low
levels for others. This will have the effect of increasing the prices of the
CFCs that are still available and processes which avoid their use are
expected to take over almost completely in the next few years. The
majority of the new plasma processes use a perfluorinated compound to
replace the CFC. and are the most commonly used.
The perfluorinated compounds (PFCs) all have negligible effects on
atmospheric ozone but they are long-lived in the upper atmosphere, from
hundreds to tens of thousands of years estimate), and have
strong infrared absorption bands. The latter give rise to global warming via
the greenhouse effect (section 2.5.12). However, emissions from semicon-
ductor etch processes worldwide is modest; only a few per cent of total
emissions. The worst polluters for perfluoropropane are electrolytic alu-
minium processing plants. Only in the event of aluminium smelting
cleaning up considerably would strong pressure fall on the semiconductor
industry to eliminate PFCs.
Despite this, there are concerns about PFCs and many wafer fabrication
plants are seeking to reduce PFC emissions, either simply by more careful
gas control, using lower pressures and shorter etches, or by actively
removing PFCs (and their breakdown products) from the exhaust gas. Wet
and dry scrubbers (section 2.5.14) can both be employed and give good
results with care. Dry absorber scrubbers can be effective but need correct
incineration disposal after use. Some of the other scrubber problems are
that, for wet scrubbing to work, a high-temperature flame burn-box is
necessary to ensure that the PFCs (or allowed CFCs) are broken down to
more reactive radicals and compounds that will be effectively scrubbed by
the wet-spray chamber. The integrated burner/scrubber submerged flame
technique is also claimed to be good here, as are dry scrubbers which are
heated, like the Edwards GRC system.
An alternative that is now edging up towards acceptable performance is
dry treatment at low pressure, i.e. before the vacuum backing pump. This
employs a plasma to excite the exhaust gas mixture so that it will react,
yielding mineral deposits which are caught in a chamber before the backing
pump. This kind of scrubber provides a useful protection for oil-sealed
vacuum pumps which soon wear out under such conditions. (Even with the
oil circulation and filtration and frequent oil changes a particulate load will
reduce pump and oil lifetime considerably.) The newer Northey claw
backing pumps also suffer from wear and occasional overloads when
operated with particulates. The pre-pump filter (necessarily a coarse filter
to maintain flow rates) or cyclone chamber will intercept many of the
particulates but will leave them as an almost unagglomerated powder,
which is easily re-entrained. The striking of a plasma causes the deposits
to become much stronger, more like eggshell in texture. Large-area
electrodes allow several kilogrammes of material to be collected by such
plasma traps.
3.5.10 Other processes in the wafer fab
Annealing processes for silicon wafers. After a number of processes to
which the wafers are subjected, point defects, small amounts of damage
done to the perfection of the single crystal, will arise that will impair the
device operation, for example by reducing carrier lifetime. These defects
can be repaired by a controlled heating process or annealing. Ion
implantation, for example, causes a lot of sub-surface damage such as
interstitial atoms and vacancies, and is invariably followed by annealing.
56
A
heating cycle may also be necessary after implantation or diffusion to carry
out drive in, that is, to diffuse dopants from near the surface into deeper
layers of the device. Similarly, the last traces of impurity from a wafer can be
removed by a controlled heating process called gettering. The heating in
this case causes mobile impurities to diffuse until they reach oxygen-induced
stacking faults in the silicon or microcracks on the back side of the wafer.
The impurities are then trapped by the faults or cracks.
Defect damage can be repaired by controlled heating at high temperature,
typically done in an inert atmosphere of nitrogen or argon. A tube furnace
can be used for annealing a whole boat of 25 or more wafers or single wafers
may be annealed in an RTP furnace.
Purging of reactors. Before a furnace or reactor operation can be
performed, the ambient air, which would react vigorously with silicon at
high temperature and with the active ingredients of process gases, must be
expelled. Frequently there is a vacuum step but this is only possible with a
vacuum-qualified reactor, i.e. one with good seals and strong enough to
withstand atmospheric pressure. This is, however, not always sufficient.
Hence there is a huge demand for purging gases, typically nitrogen. The
purging gas requirements are difficult to estimate since the purging process
itself is not always predictable and the effectiveness of the purging required
will vary with equipment and the type of ICs being made.
For ordinary (and now increasingly old-fashioned) tube diffusion
furnaces, purging is at a rate of a few litres per minute or so, rising during
loading and unloading to provide additional shielding of the wafers from
atmospheric contamination and to provide additional cooling. While not in
use, purging must be maintained at some level to keep the furnace ready for
56
Note that annealing of silicon is not done to soften the metal, as in metallurgical annealing of
steels, aluminium or copper. Annealed silicon remains the same as the original cut wafer, i.e. a
bright silvery metallic-looking material that is, in fact, brittle.
use. Once air has been let in, the system will need to be purged for hours and
baked at temperature before it is ready again.
Gas ballast for vacuum pumps. Pyrophoric gases such as silane require
dilution with nitrogen before they can be diluted with air and expelled up the
factory chimney. Similarly, vacuum pumps used on many processes are
subject to problems because of the condensible nature of the gas load they
are evacuating. The recommended solution is the addition of a small flow of
nitrogen to the final stage of a multistage rotary vacuum pump. This is
known as gas ballast. Manufacturers will advise on the flow rates needed
and the (small) loss of ultimate vacuum this imposes.
Note that the new types of dry vacuum pumps can take larger flows of
ballast than oil-immersed types. Dry pumps are those recommended for
applications where there are corrosive condensible vapours. A gas ballast
will almost certainly be essential with these gases.
Blow-drying. The blow-drying of wafers after a wet clean or wet etch
process is a common application of nitrogen. In principle, high-quality clean
dry compressed air (CDA) could be used instead and this is the practice in
some wafer fabs. The use of a blast of dry nitrogen to evaporate solvents
from newly coated photoresist is also common for both convenience and
quality assurance. The inert nitrogen may avoid possible problems with
resists that are more sensitive to oxidation.
Vacuum wands, blow guns and pneumatics. Vacuum wands are venturi-
operated paddles which are used whenever it is necessary to handle wafers
manually. They allow handling of wafers via the back surface only and avoid
touching of the delicate edge or patterned surfaces of the wafer. Blow guns
are handheld valves supplied with nitrogen via a filter and used to blow-dry
or clean-off visible dust particles from wafers, again where manual handling
is necessary. Pneumatics, which would normally be operated by oil-mist
lubricated compressed air, are often operated in a wafer fab by nitrogen;
certainly the oil-mist in standard compressed air is totally unacceptable in a
wafer fab. These are all applications for a clean dry gas and nitrogen is
frequently employed, again with some companies using CDA where this is
available.
Gas inerting for IC storage. For storage and transport of the chips, they
need to be kept in a clean, dry and preferably inert atmosphere. Nitrogen is
the gas of choice and considerable amounts can be used where chips are
stored in large banks of cabinets, all of which need to be kept freshly purged.
Practically all glovebox handling cabinets for semiconductors (sometimes
called dry boxes) are purged with nitrogen.
Chemical handling. Even the use of liquid rather than gaseous processes
does not mean that no gas is used. Frequently nitrogen is used for propelling
corrosive liquids without the use of pumps (pressure transfer or pneumatic
lift pumping) and for agitating the reaction baths. Even where centrifugal
spray processing is used, employing liquid reagents, considerable quantities
of nitrogen may be consumed.
Point-of-use chemical generation. Aqueous chemicals, such as hydro-
chloric acid, ammonium hydroxide, and hydrofluoric acid, are difficult to
prepare in ultrapure form. They tend to pick up particulates and metallic
ions during their preparation and more contaminants are added during
transfilling into transport containers and pumping to the point of use from
those containers. The very fact that they are used to clean-off metallics and
particulates from wafers should tell us that this is to some extent inevitable.
These tendencies can be almost, but not completely, suppressed by the use
of special materials. Containers and pipelines made of PVDF (poly-
vinylidene fluoride), PFA (perfluoralkoxy) and high-purity PE (polyethyl-
ene) polymers are, for example, favoured. However, even these highly inert
polymers contain residual catalyst, reactive end-groups, etc., which can
result in slight contamination.
The ultimate way to control all this might be the preparation of aqueous
chemicals by dissolving the appropriate ultrapure gas in ultrapure water at
the point-of-use chemical generation (POUCG). The ultrapure water
needed is already available at a wafer fab where water is treated by ion
exchangers followed by reverse osmosis to achieve very high levels of
purities. Resistivity very close to the theoretical maximum of around 18.26 x
is normal in wafer fabs, indicating that the ionic content of
cleanroom deionised water is very good (just 10 ppb salt will reduce the
resistivity of ultraclean deionised water by half). The gases used, HF, HC1
and NH
3
, can all be filtered to remove particulates down to or better
and cannot intrinsically harbour metal ions so chemicals generated in this
way have lower levels of metal ions and particulates.
Several companies are now offering this approach using equipment which
will prepare small batches of chemical near the wet bench or other
point-of-use inside the wafer fab. An ultraclean oxidative aqueous solution
to supplant hydrogen peroxide can also be provided by this approach, by
bubbling clean ozone, made from ultrapure oxygen, through water. Ozone
solutions are problematical, however, as they have a very short lifetime and
concentrated solutions are not possible.
Although the technique is in its infancy, a typical installation is already
becoming evident. In this a tank containing a few litres of deionised water is
gassed with ultrapure anhydrous gas, which has been filtered to remove
particulates. Two or more gases can be added to form ready-to-use mixtures
used in wafer cleaning and etching processes. An example would be the
buffered oxide etch (BOE) used on silicon dioxide layers, consisting of
aqueous ammonia and HF. Another example would be an aqueous
ammonia/ozone mixture (with hydrogen peroxide instead of ozone
this would be SC1 or RCA1). The same apparatus can then be used to
generate the mixture which often follows, i.e. an HCl/ozone mixture (with
this is would be SC2 or RCA2). Ozone mixtures are not yet used
routinely, however, because ozone cannot be delivered in simple gas
cylinders but must be generated by a relatively expensive electrical
discharge generator (section 2.2.7).
With an increasing move towards more dilute cleaning solutions, this
approach is going to look increasingly attractive, despite generally falling
prices for ultrapure chemicals delivered. Already, POUCG equipment is
claiming <50 ppt levels or better for all metal contamination, comparing
favourably with the ppb levels typical in many high-quality delivered
chemicals.
3.5.11 Gases in IC assembly and hybrid circuits
It is not simply the wafer fab sector of the electronics industry that uses
industrial gases today. The IC assembly process, in which silicon dice are cut
from wafers and put into their protective ceramic or plastic jackets, is also
using industrial gases, although not with the same order of purity require-
ments. The manufacture of hybrid circuits also uses gases in modest
quantities.
Many applications in the manufacture of hybrids and packaging use
nitrogen as an inert atmosphere. Great care must be taken in these
applications to ensure a low oxygen/humidity environment throughout, as
levels of only a few hundred vpm (volume parts per million) can cause
problems that are notoriously difficult to trace. The nitrogen itself is pure
enough for any of these applications. However, problems may arise with
inefficiencies in the purging process, perhaps due to backwaters in the flow
path or simply due to the purging flow rate being too small.
Once wafer processing is complete, hundreds of IC circuits are formed in
the surface of each wafer. Subsequent operations to make them usable in
actual operational equipment are required; these are collectively known as
packaging or back-end. The first operation is wafer probing. Clusters of
tiny spring-loaded probes are lowered onto each circuit on the wafer and a
low bandwidth basic test carried out. The second operation is to saw the
wafer into individual circuits using a high-speed diamond circular saw. The
individual naked circuits (known as dice or, colloquially, chips) are then
sorted and the dice which were bad at probe are rejected. Dice may then be
further tested, although it is by no means easy to connect them to and
disconnect them from a high bandwidth testing circuit on a high speed
Figure 3.27 Dual in-line packaging: (a) side view; (b) top view; (c) cross-section of ceramic
package.
production line. Dice which pass these more stringent tests are known as
known good dice or KGD. KGD are often required in hybrid circuits,
where the failure of just one die can lead to an expensive rejection of the
whole hybrid with all its dice.
IC packaging. Dice may then proceed either by the plastic packaging or
the hermetic packaging route. In the former, the naked ICs are bonded
(the die-bonding process) to a metal frame which will become the
completed packages leads, i.e. a lead-frame. Die-bonding may be by
silicongold eutectic solder or conductive epoxy. Individual parts of the IC
are then connected to the metal leads of the frame with fine gold wires
(wire-bonds) attached by a microwelding process. The whole assembly,
leaving the leads protruding, is then encapsulated in epoxy in a moulding
machine and the complete package cropped to separate it from the
lead-frame edges. In hermetic packaging, the ICs are bonded down to a
metal or ceramic package then wire-bonded to leads sealed into the package
by the package manufacturer (see Figure 3.27). A lid is then added under
moisture-free atmospheric conditions, either by welding or by reflow of
low-melting-point glass. By warming up the assembly to 150 or 200C during
packaging, much moisture can be driven off, ensuring freedom from
corrosion effects. Many of these packaging processes are carried out in dry
boxes in nitrogen gas to ensure that moisture is not spuriously included in
the packages, to avoid discoloration of welded joints and to avoid oxi-
dation damage to the ICs caused by hot processes. Silicon chips are suscep-
tible to oxidation in room air but oxidise especially quickly above about
100C.
Hybrid circuits. Manufacturers of hybrid circuits are also users of gases.
Hybrid circuits are so-called because they are a hybrid between integrated
circuit technology and printed circuit boards. They also lie in between
monolithic silicon integrated circuits and printed circuit boards in their ef-
ficiency at packing in circuits. They typically comprise a circuit board,
several times smaller than a conventional glass-epoxy PC board, made of a
ceramic such as alumina. (Other hybrid systems are based on silicon sub-
strates using techniques more closely related to the monolithic silicon IC
industry. ) In alumina-based hybrids, the alumina is the substrate onto
which simple electronic passive components such as capacitors or resistors,
as well as conductor patterns, can be printed using special inks which can
be fired at high temperature. The inks are composed of a frit or powder-
glass-type material, together with palladium or gold for conductive inks,
ruthenium dioxide for resistors and high dielectric constant glass for ca-
pacitors. Once printed on the substrate, the inks are fired at temperatures
from 500 to 1000C, when the frit component flows, binding the active
layer together and bonding it to the substrate. After several printing and
firing cycles, the board part of the circuit is complete and it remains only to
screen solder paste onto the board, add active components such as ICs and
transistors in special miniature surface mount (SM) packages, and reflow
the solder. The whole circuit will often then be encapsulated in epoxy. This
is essential if instead of SM ICs, unpackaged naked silicon dice are used.
These are wire-bonded onto the hybrid or inverted and joined by heating,
causing reflow of solder balls deposited on them to effect electrical connec-
tions. The latter process is called flip chip because the solder balls are on
the active surface of the silicon so the silicon chip must be turned upside
down to bond them on.
The inks used for alumina hybrids have to contain organic binders to
make them flow and handle in the screen-printing machinery typically used
for patterning. The ink must flow sufficiently well to be easily squeezed
through the mesh of the screen but not so runny that it bleeds, i.e. flows
away from the desired patterned areas, after deposition. The organic
binder is burnt away early in firing leaving the glass frit to bond the ink to
the substrate and the metal or oxide to form the active part of the layer.
Many hybrid circuits use air-firing ceramic inks, such as palladiumsilver
conductor ink and ruthenium oxide-based resistor ink. However, many cir-
cuits are now being switched to more highly conductive inks based on
copper, which cannot be air-fired. To go with these, new resistor and in-
sulator formulations have been devised so that the whole hybrid can be
fired in nitrogen. Copper-based hybrid inks will also oxidise far below firing
temperatures and many should be heated in nitrogen even at the low
temperatures first used to burn out the organic binders.
Fault detection and thermal testing. Intermittent faults due to faulty
contacts or semiconductors in circuits are a serious form of failure and can be
frustratingly difficult to locate for repair. It is standard practice in electronic
fault-finding to spray suspect components with a liquefied gas. On cooling to
50C or so an intermittent bad connection will often fail, allowing it to be
traced. A non-toxic, non-inflammable, non-conductive liquefied gas boiling
at 50C or so, which can be stored at room temperature in a cheap
disposable spray can, is needed. With CFCs such as CFC-12 banned for
applications of this sort, the choice revolves around materials such as
HCFC-22 or HFC-134a or mixtures of these.
To find faults early, and to assure functioning at all temperatures (for
military-grade ICs), new ICs and complete circuit boards or other subassem-
blies are often tested in environmental test chambers, typically in the range
These chambers are required to be able to reach and then
maintain any desired temperature within minutes. Often thermal cycling
tests are needed, in which many cycles of heating and cooling must be
completed in the minimum of time, and the powerful cooling effect of a
cryogen injection system is often used to achieve this (section 3.7.3).
Lid brazing. Many ICs and hybrids are packaged in metal-plated ceramic
packages designed for metal lids. Metal lids, typically gold-plated Kovar,
are then fixed to the ceramic or metal header by a brazing or welding
operation. Despite the plating, the packages and their lids would be
damaged by air, as would the IC connection pads and wire bonds if they got
hot enough, and it is necessary to maintain a nitrogen or argon atmosphere.
Ceramic lids are melted down onto a low-melting-point glass layer on the
package body. A nitrogen atmosphere is maintained in order to ensure that
the headspace above the chip will not contain moisture or oxygen, which
would attack the chip over a period of time. Inert atmospheres are not just
used in ceramic IC sealing; there are also advantages in using nitrogen in
some plastic encapsulation processes.
Helium leak testing. The hermetic seal of packages may be checked by a
helium leak test. There are many ways of doing this but one simple way,
requiring no particularly expensive equipment, is as follows. Packages are
placed in a chamber and pressurised with helium at a few bar pressure. They
are then removed and placed in a bath of solvent. Packages with leaks will
have taken up He and become pressurised, and on placing them in the bath
at 1 bara, small bubbles of He will be seen, even in the case of quite small
leaks. More sophisticated systems employ mass spectrometric detection of
He.
Nitrogenhydrogen mixtures. These can be employed wherever a high-
temperature operation may cause undue oxidation of silicon or other metal
surfaces, even with nitrogen inerting. These forming gas mixtures are
especially helpful where small amounts of oxygen cannot be avoided in the
high-temperature zone. Although often used in the wafer fab itself, these
mixtures are more often encountered in electronics assembly; for example in
heating ceramic packages to avoid oxidation of the metallised surfaces prior
to lid seal or die-bonding.
3.5.12 Gases in compound semiconductor manufacture
Most semiconductor devices are made of silicon. However, devices for RF
and microwave regions and devices such as light-emitting diodes, use other
semiconductors. A few per cent of the semiconductor industry is based on
the alternative IIIV and IIVI compound semiconductors such as gallium
arsenide and indium phosphide.
GaAs is a much more difficult material to work with than silicon and other
compound semiconductors are more difficult still. One of the biggest
difficulties is that there is no simple process for an insulator. With silicon,
silicon dioxide is an insulator that can be grown on or added to silicon and is a
reasonable thermal match to it. No such convenient insulating oxide can be
grown on GaAs and deposited layers are also fraught with difficulties.
Another difficulty is that single GaAs wafers are difficult to make with the
necessary levels of purity and freedom from faults. During growth, the more
volatile arsenic tends to be lost from the melt, giving rise to vast numbers of
faults if uncorrected. Epitaxial layers are often therefore needed in GaAs
but they are more difficult than for silicon; a mixed gas atmosphere is needed
with gaseous sources for both arsenic and gallium which will deposit pure
GaAs. Trimethyl gallium and arsine are the standard gases for
carrying this metallorganic vapour phase epitaxy (MOVPE) process. The
process is sufficiently tricky that some device manufacturers do not use
MOVPE themselves but buy in wafers with MOVPE layers already added to
their specification. There are companies which specialise in carrying out
GaAs MOVPE as a service in Europe and the USA.
Doping and alloying of GaAs and other compound semiconductors is
accomplished with unusual compounds, such as dimethyl zinc, dimethyl
cadmium, disilane germane hydrogen selenide, hydrogen
telluride, tripropyl aluminium, trimethyl aluminium, trimethyl indium,
tetramethyl tin, stibine and other metallorganics. Only minute amounts of
these compounds are needed; the entire world market for all these
compounds is only a few hundred kilograms.
Most of these gases are thermally unstable, that is, they are endothermic
and release heat on decomposition. Instability leading to loss of product can
be avoided by dilution with, for example, hydrogen; this is standard practice
for diborane. Another problem of endothermic gases is that they can
actually undergo an unwanted CVD process inside the gas cylinders they are
supplied in. Occasionally a cylinder valve on these gases will fail because
deposition has occurred on the valve seat and they cannot be turned off
completely, giving rise to a dangerous leak. An additional technique which
is being looked at seriously is that of using the deuterium isotope to form
hydrides. Fully deuterated diborane has enhanced stability, for example. As
noted above (section 2.2.10), deuterium is the least expensive of the
separated isotopes and is not an unreasonable cost for gases for GaAs use,
where quantities used are in tens or hundreds of grams.
Compounds such as germane and hydrogen telluride are sufficiently
endothermic that a spark or other source of ignition can start a runaway
chain reaction in which the heat released from the decomposition is
sufficient to make surrounding molecules to decompose, too. To avoid
disasatrous detonation, gases like germane have therefore to be stored at
relatively low pressure gas (10 bar or less), as well as being diluted with
hydrogen in many cases.
Recently, the preparation of these exotic compounds at hitherto unheard
levels of purity has begun to be carried out commercially by companies such
as Voltaix and Epichem. The higher purities available will enable the
preparation by CVD and MOVPE of exotic semiconductors with properties
(defect densities, carrier lifetimes and so forth) which are closer to the
near-perfect properties now achieved only with silicon. Impurities down to
< 1 ppm of metals are now routinely achieved, whilst analysis with ICP-MS
down to ppb levels can be carried out for critical impurities. Even more
sophisticated analytical chemistry tools are now being brought to bear on
occasion. For example, 250 MHz NMR can see oxygenated compounds in
trimethyl aluminium; this is important in final devices.
The Epichem company (Bromboro, UK) is an afficionado of the adduct
purification route. Adduct purification is only justified for purer grades; up
to 99.5% purity alkyl compounds are straightforwardly prepared using
standard materials and techniques. It is only beyond this purity level that
another technique is needed. An example of adduct purification is the
preparation of ultrapure trimethyl aluminium.
Bis-diphenylphosphenoethane DIPHOS, in which
Ph is is first reacted with trimethyl aluminium (TMA). The resulting
adduct compound, DIPHOS-TMA, is an involatile solid that will contain
only adducts with trivalent compounds, the adduct bond being a donation of
a pair of electrons from each of the phosphorus atoms in the DIPHOS.
Impurity compounds will not form adducts and can be extracted by vacuum
and gentle heating as they are mostly volatile. Heating to 100C will
decompose the adduct compound, releasing the TMA and allowing a further
stage of distillation. Similar reactions are used to purify the other trivalent
organometallics such as trimethyl gallium and trimethyl indium.
These compounds are now used in MOVPE growth. Wafers up to 6 in
diameter, based on GaAs and InP, can now be made with high-purity
epitaxial layers and precise levels of doping. Lasers and high efficiency
LEDs are now made from gases prepared in this way, as they show much
better performance. GaInP LEDs made by MOVPE seem likely to replace
liquid-phase epitaxy devices as they are up to a thousand times brighter with
higher electrical efficiencies.
3.5.13 Gases for optical fibres
Optical fibres are an example to rival silicon semiconductors of an industrial
development which depends on an ultrahigh purity material, albeit on a
smaller scale. Optical fibres are the principal means used to conduct
long-distance telecommunication traffic around the world. Electric signals
from telephones or computers are converted into pulses of light and these
light pulses are directed along the centre of a solid fibre of quartz glass which
acts as a light pipe.
Light approaching the interface of a piece of glass with a medium of lower
refractive index from within at a shallow angle is reflected. In fact it
undergoes total internal reflection; 100% of the energy is reflected back into
the glass. This principle is the reason why a piece of glass fibre can be used as
an efficient light pipe. Every 50 or 100 km of optical glass fibre there is a
repeater station, which amplifies the weakening light pulses and passes them
on. At the receiving end, the light pulses are reconstituted into electric
signals for the receiving telephone or computer.
Ordinary glass even a couple of feet thick noticeably tints and dims light
passing though it. Workable optical fibres are thus wholly dependent on
achieving the utmost clarity (and therefore purity) in the glass used.
In fact, many of the processes and gases pioneered in the semiconductor
industry were taken over and modified to suit the manufacture of optical
fibres. Helium, nitrogen, oxygen, silane, HC1, and
chlorine are all used in the industry.
The process of producing an optical glass fibre bears some resemblance to
the process used since the 1800s for the preparation of rock sweets (sticks of
mint candy with a name or picture running through their cross-section). The
required fibre has a high optical density core of in a body of
lower optical density. To achieve this radial variation of optical refractive
index, a large, short, fat boule or preform is produced that has a much
larger (several inches) diameter. The boule then heated and drawn down
from a few inches to 100 m in diameter by heating until it reaches the
Figure 3.28 Glass preform lathe arrangement.
consistency of sticky toffee and then pulling out a sticky string of glass on to a
carefully controlled winding drum (Figure 3.29).
The preform can be produced in many ways but must maintain extraordi-
nary levels of purity. One process uses a CVD technique to deposit a soot
of porous glass onto a central silica rod. The central bait rod is rotated in a
lathe machine whilst a burner is played onto it, moving laterally to ensure an
even distribution of deposit along the axis. The burner is fed with methane,
oxygen, and dopant chlorides; first heavy dopants such as then
lighter ones such as The radial variation of composition and refractive
index of the fibre will be determined by the programmed changes in the
flame dopants during the deposition of the preform (see Figure 3.28).
The reactions in the flame are analogous to the CVD reactions used in the
semiconductor industry to lay down the final insulating glassivation layers
on silicon but also have some similarity to flame-coating processes since a
fine particulate soot of hot glassy material is formed in the flame before
deposition.
The presence of any moisture in the process would leave OH radicals in
the glass and ruin the fibre so conditions of high cleanliness are needed and
gas inerting is used in subsequent processing. However, the flame process
adds a lot of moisture to the boule and this must be eliminated. First the
central rod is removed. The resultant hollow porous dry glass rod is then run
through a tube zone-sintering furnace to convert it into a solid boule. The
use of a helium atmosphere in the sintering furnace enables bubbles to be
more easily eliminated. Heating in hydrogen chloride or chlorine gases,
sometimes in mixtures with oxygen, is then used and results in the
elimination of residual moisture and OH.
Other preform methods are closer to semiconductor CVD as they form a
Figure 3.29 Fibre preform melting and pulling machine.
deposit on the inside of a glass tube that is heated by a flame from the outside
and do not need to add methane fuel gas to the deposition mixture, with
consequent moisture problems. As in the above method, the tube must be
furnace-treated to eliminate the central hole and convert it into a solid
boule.
The production of fibre is done in a pulling tower by pulling a gob of
molten glass from a hot boule, suspending it in the furnace, then stretching
this to form the initial fibre (Figure 3.29). The fibre is wound onto a take-up
spool, whose speed is controlled to control fibre diameter, and continuous
production can begin. Once the machine is running, at up to a few metres per
second, the fibre is cooled within a fraction of a second of its production and
coated with a plastic coating which protects it from crack-forming or surface
contamination. Coated fibres are very strong, despite being only or
so in diameter, which makes handling easier.
The pulling furnace may be a small oxy-hydrogen torch furnace, a
graphite tube furnace or another type but is invariably a vertically mounted
tube furnace. Ideally an inert atmosphere is maintained around the fibre in
the furnace. Cooling the newly formed fibre in a pure gas preserves its
perfect surface and increases the potential production rate of the fibre. An
inert gas also means that a higher quality UV setting lacquer (without
additives for counteracting air oxidation) can be used for coating. Even with
thermoplastic coatings, an inert gas for the coating section can avoid some of
the thermal degradation of the plastic that may occur in air.
3.5.14 Gases in general electronics
Soldering. Although traditionally done in air, soldering of electronic
components to printed circuit boards, using nitrogen or other gases, is now
becoming more common. Solder is usually a joint between an electronic
component, lead-wire (made of copper, often coated with tin or solder), and
a circuit board (also copper, often tinned). The solder used is usually a
simple 62/38 tin/lead eutectic composition, which melts at 183C, although
small amounts of additional alloying elements such as antimony can give
solder melting at a little higher or lower temperature, allowing multistage
soldering. Multistage soldering requires that subsequent soldering does not
loosen previously soldered components. A little flux is employed to aid
wetting of the joint by the molten solder. The flux is a mildly acid material
such as ammonium chloride (a powerful flux), adipic acid (mild) or a rosin (a
mixture of natural diterpene acids with a milder action, favoured today).
The flux removes a slight oxide film and protects the soldered surfaces from
further oxidation during the heating that immediately precedes the molten
metal.
The two most common techniques for soldering (other than by hand) are
reflow and wave soldering (Figure 3.30). After placement of the com-
ponents, sometimes using glue to retain them, the latter process passes the
circuit boards through a wave of molten solder. In the reflow process, the
components are placed/glued after screening a pattern of solder paste onto
all the places where the component will be. The boards are then passed
through an oven where the small patches of solder paste fuse into soldered
joints. (Another gas soldering technique uses a hot blast of nitrogen for
fusing solder paste to fix small, single components.)
Nitrogen, used in wave soldering, reduces the amount of dross and
improves quality, although it is still unusual to see gas inerting on
wave-solder machines. Dross is a mixture of solder and oxides of lead and tin
that accumulates on the surface of molten solder after a few minutes
exposure to air. Dross is a considerable problem, causing bridging and
wetting problems on circuit boards, and often requiring manual intervention
in the automatic soldering process to remove it (it can be reprocessed back
into solder). A large wave-soldering machine may form more than 1000 g
per hour of dross, on a solder wave just 20 cm across. With small reductions
Figure 3.30 Wave soldering with gas inerting.
in oxygen content, dross is produced as copiously as ever. However,
reducing the oxygen below about 2% in the atmosphere makes improve-
ments and dross formation is minimal at 0.1% or less oxygen.
Gases have not hitherto been much used in reflow soldering but this may
eventually change. Traditionally, circuit boards that have been soldered go
through a cleaning bath based on CFC liquid solvents to remove residues of
the flux used and ensure wetting of the metal parts by the solder. This is no
longer environmentally acceptable. New approaches vary from the use of
different solvents (one new solution is a soap and water system), to
avoiding the creation of a flux residue in the first place.
The solder paste used in reflow is a mixture of powdered solder, solvents
and flux. The use of nitrogen inerting in the reflow oven enables the use of
milder fluxes, in smaller amounts, allowing circuit boards to be used in less
exacting applications without cleaning. One process was developed using a
flux which reacts with the gas used in the reflow oven, leaving an extremely
small residue. No-clean fluxes are now available from several suppliers that
have extremely low residual deposits and do not need cleaning-off. No-clean
fluxes are based on very mildly acidic compounds, such as adipic acid, and
have a small solids content, relying on the use of nitrogen to largely protect
the copper or tinned surface in the hot atmosphere until solder wetting
occurs. In air, they would not be able to stop excessive oxidation and
inadequate wetting occurring.
Occasionally, references are made to vapour-phase soldering. This is a
form of reflow soldering in which vapour from a boiling liquid, contained in a
deep t ank, condenses on a circuit board lowered into it and heats it up to
near the liquids boiling point. Inert fluorocarbon liquids that boil at 300C
are used for this process. As a technique it is now regarded as obsolescent,
although it does provide some of the advantages of an inerted soldering
technique.
Gases in printed circuit board manufacture. The preparation of multilayer
printed circuit boards (PCBs) is now a sophisticated technology. The use of
nitrogen-filled autoclaves for laminating them has become a standard
procedure.
The laminae, green epoxy/glass fibre composite sheets a fraction of a
millimetre thick and patterned with copper tracks, are placed together in
stacks and placed inside a vacuum bag. Hundreds of such bags are then
connected to a vacuum pipe inside a large heated pressure vessel (an
autoclave). Whilst a vacuum is drawn inside the bag, a 10 bar nitrogen
atmosphere is applied and the whole apparatus heated. After a few hours,
the epoxy resins are set and the boards can be removed, the laminae
efficiently bonded together. The use of nitrogen avoids risks of fire and
explosion that the use of high-pressure air with the large volumes of resin
would entail.
PCBs, after lamination, are drilled and the holes are plated with copper
using an electroless copper process followed by electroplating, providing a
connection between each side of the board. After drilling the PCBs, it is
often the case that the sides of the holes are smeared with resin, preventing
the correct plating of plated-through holes. The holes can be desmeared by
processing in an oxygen/argon plasma (section 3.5.7), which quickly oxidises
the resin blobs away while leaving the bulk board more or less unaffected.
3.5.15 Thermionic valves
Although semiconductors now account for 99% of all electronics sold, there
are some applications for which the thermionic valve is still the only
solution. The cathode-ray tube (CRT) used in television sets and in
computers is probably the most common thermionic valve today. In the
CRT, the display seen is simply the result of fluorescence caused by
electrons from the hot filament striking the screen. Magnetrons for
microwave ovens are probably the second most common thermionic device
in use. Similar but larger thermionic valves, up to 2 m long, are used for
television and radio broadcasting, and many other high-power RF appli-
cations. All these devices employ a similar range of gases in their
manufacture but magnetrons, still made in large quantities today, are a good
example. The quantities of gas used are not necessarily as small as might be
thought; one magetron plant, for example, specified 50
tonnes of Ar, and 1000 tonnes of LN per annum.
Magnetrons consist of a glass envelope with a filament and various metal
electrodes inside, and a powerful magnet around the outside (Figure 3.31).
They create powerful microwave electromagnetic waves in the 1 to 10 GHz
range, typically 2.4 GHz, which can penetrate several centimetres into
non-conducting objects such as food and heat them up; hence the application
for microwave cooking.
The anode (outer electrode) is typically made from copper, whilst the inner
electrode is typically a thorium (and other reactive metal) oxide-coated shell
with heater wires inside. The glass and ceramic envelope must be very
carefully sealed to the metal components. This often involves furnace
processes to form a controlled oxide film on the metal prior to sealing. The
sealing process itself can be performed in a nitrogen/hydrogen atmosphere
belt furnace.
Glass-to-metal and ceramic-to-metal seals are in principle difficult but have
long been studied and many solutions are now available. Molten glass bonds
well to ceramic and to low expansion metals such as Kovar (Fe/NiCo 54/29/17)
when the metal has a controlled thin oxide film on its surface. Metal powder
mixtures such as Mo/Mn can form a sintered bond between ceramic or glass
and metal when heated to 1400C in a hydrogen/steam atmosphere. Ceramics
may be sealed to metals by brazing or soldering. In order to do this, the
ceramic must first be coated with a metal. The metal plate may be achieved by
sputter-coating and then electroplating to form a thick, well-bonded metal
surface. Alternatively, a paint composed of a hard metal powder in a
nitrocellulose lacquer can be painted onto the ceramic parts. If fired in a
hydrogen/steam atmosphere, paints of this sort will form a solid metal layer
for brazing.
The magnet of a magnetron is typically a permanent type made from a
dispersion type of FeCo/NiAl (Alnico) cast alloy, quenched and then
annealed at 500700C. Sometimes clever techniques involving cooling
(under a controlled atmosphere) from a molten mixture in a magnetic field,
which allows the full potential of high performance alloys, such as Ticonal
and Alcomax, to be realised. Alternatively, barium hexaferrites may be
employed (section 3.2.8). These are made by sintering a magnetic oxide
ceramic mixture in a controlled atmosphere following pressing in a magnetic
field.
3.6 Gases in water treatment
Wastewater treatment is commonly classified into primary, secondary and
tertiary treatment.
Primary treatment. This is the removal of particulate matter. Following a
screen to eliminate pebbles and other gross particulate above 20 mm or so,
settling tanks are employed in which much of the suspended solids simply
settle out to the bottom. Aiding this process is very commonly a flocculation
system. This adds a highly charged ion, such as are given by iron(III) chloride
or sulphate or aluminium sulphate, which acts as a centre for the
agglomeration of the unfilterably small colloidal particles into macroscopic
floc particles which will settle or are filterable. Sometimes specially designed
polyelectrolytic chemicals (charged organix polymers) are added instead to
achieve the same effect. Occasionally, lime or another alkali is also added to
ensure that the pH, which should not be too low, is correct. The flocculating
chemical added is often added at a constant rate. It is only needed in very small
quantities.
57
Primary treatment reduces the biochemical oxygen demand
(BOD), the amount of (undesirable) biologically oxidisable material in the
water, by 50% or more.
Secondary treatment. This is today considered an essential part of the
wastewater treatment process. In this, colloids not removed in primary
treatment, dissolved organics and other oxidisable contaminants are oxidised
by processing with colonies of microorganisms to harmless molecules such as
and water. In this way, BOD can be reduced to 10% or so of input values.
Two systems, activated sludge and trickling filter, are in use, with the former
tending to become more popular. Activated sludge simply blows air or oxygen
through the pre-treated wastewater, to which some of the sludge extracted
previously has been added (this contains the necessary dose of micro-
organisms). Trickling filters are beds of rock chippings over which water is
trickled, the microorganism colonies adhering to the chippings. Secondary
treatment, finishing with filtration, largely eliminates organisms harmful to
humans and animals, although a further sterilisation step with chlorine or
ozone is used for drinking water. An alternative to the aerobic secondary
treatment is anaerobic treatment, which, in addition to purifying the water,
can produce useful amounts of methane fuel gas. This can be used to run a gas
turbine electrical generator to power sewage plant operations.
Tertiary treatment. This is aimed at reducing non-BOD contaminants,
inorganic phosphorus, nitrogen, etc., which, although not pathogenic, cause
problems in water courses by allowing excessive growth of algae and water
plants that can clog or pollute rivers and streams just as effectively as
quarrying or detergent discharges. Addition of lime (calcium hydroxide) or
aluminium hydroxide is now used at many plants to reduce ionic content.
Water purification. This refers to operations downstream of wastewater
processing. Wastewater processing aims simply at producing water fit to put
back into rivers and streams. Water purification is aimed at producing purified
piped water for human purposes, either domestic or industrial. For piped
mains drinking water, little extra processing is done except additional
filtration if necessary and sterilisation of any remaining pathogenic
57
There is now a patented system (Matchfloc Ltd) that will regulate the amount added.
microorganisms by chemical oxidation. Chlorine, chloramine
produced by injecting ammonia into previously chlorinated water), chlorine
dioxide and ozone are all used for sterilisation. However, ozone may well
become the dominant sterilant in the future (section 4.3).
Water softening. This refers to the removal of metal ions. For many
industrial purposes, it is necessary to ensure that water is soft, that is, it does
not contain more than a minimal amount of calcium and magnesium salts,
which deposit as hard, irremovable scale when water is evaporated or heated.
Hard water seriously limits the effectiveness of washing with soaps and
detergents, and so soft water may also be regarded as desirable for domestic
purposes. However, soft water is now regarded as undesirable for drinking,
being associated with increased rates of coronary heart disease. In the older
developed world lead pipes have been installed for drinking water, only
rather recently being replaced by copper and plastic. There are thousands of
kilometres of lead pipe still in existence, which release lead ions slowly into
soft water, whilst transmitting hard water uncontaminated. In areas having
naturally hard water, therefore, those where water comes via limestone or
chalk rocks, for example, there is a strong move away from municipal water
softening and towards leaving hardness in water pipelines (or even adding it,
see section 3.6.2). Water softening apparatus is now usually employed at the
point of use. Water softening is carried out in general by ion exchange resins,
which replace calcium and magnesium ions by sodium ions taken from brine.
Beyond water softening, further purification processes may be carried out,
e.g. deionisation (in which ion exchange treatment is carried further to
remove both anions as well as cations), distillation (this is rare nowadays,
except on a laboratory scale or when it is carried out under reduced pressure
with a low-grade waste heat source), reverse osmosis (to remove salts
completely), UV treatment (to remove residual organics) and degassing or
gas stripping.
3.6.1 Oxygen for controlling aerobic sewage treatment
After use in domestic processes water is contaminated, mainly with
suspended fine mineral particles, salts and organic compounds ranging from
animal and human excreta to soap emulsions (from washing processes), and
often including appreciable concentrations of bacteria and other micro-
organisms, some of which are harmful pathogens. Industrial water uses may
add to these contaminants water of too high or too low pH, dissolved heavy
metals and non-biological organics such as solvents. The most serious
problems in terms of volume derive from domestic biological contaminants,
however, and it is to these that the most effort of wastewater treatment is
aimed, principally through aerobic biological treatment.
The primary measure of success of secondary wastewater treatment is
usually the reduction achieved of the BOD of the water being treated. BOD
expresses the level of undesirable organics in the water in terms of the
oxygen that will be needed to oxidise it completely. Typical BOD figures are
difficult to specify, being highly variable between sites and seasons, but
figures up to around are often encountered.
Aerobic sewage treatment requires a high level of oxygen gas to be
dissolved in the wastewater being treated. However, oxygen has low
solubility in water and this sets a large minimum size for sewage plant. This
minimum size is made even bigger if air must be relied on as the source of
oxygen. The partial pressure of oxygen in air is only 0.21 bara, giving a
maximum equilibrium concentration of about , i.e. 1 part in
With pure oxygen the maximum concentration achievable in shallow tanks is
simply because of the higher partial pressure. This would in
principle allow pure oxygen sewage plant to be five times smaller in volume
relative to air plant.
Although air dissolution is improved by using a deep or pressurised
injection point, it is still a fundamental limitation that the concentration
achievable is very low. There is another problem, the gas stripping effect.
The effect arises because the nitrogen content of the air acts as a sparging
stream, stripping malodorous volatile materials from the wastewater. If
pure oxygen is used and carefully regulated, then virtually all the oxygen is
dissolved, with the result that no gas-stripping of malodours occurs.
Measurements at plants with bad odour are occasionally made and levels of
up to 20ppm , for example, can often be found. At these levels is not
simply unpleasant, it could be dangerous to operators of the plant.
Using oxygen injection, oxygen levels higher than saturation can be
achieved for a few minutes; 200% supersaturated water will only slowly
generate microbubbles and these take some time to coalesce and rise to the
surface. This effect can be useful where reaction or dilution will allow the
oxygen to react or dissolve within a short time and can substantially increase
the amount of oxygen it is possible to deliver over the maximum amount
calculated using oxygen equlibrium values.
The systems for injecting oxygen into wastewater basically fall into three
categories:
1. inject fine bubbles into the bottom of the processing tank;
2. pump a stream of sewage from the tank into a pressure vessel where
oxygen is added under pressure, then return the stream to the processing
t ank;
3. add a stream of oxygen to the sewage whilst it is inside a high-pressure
pipeline.
Method 3 can be very low in capital but depends on a suitable injection
point. If the injection point is low in a system, and thus at a high hydrostatic
pressure, or is after a pressure-raising pump, simple injection into the water
can be very effective. The high pressure of the water means that the oxygen
solubility is increased roughly in proportion to the absolute pressure, and
ensures rapid and complete dissolution of the oxygen. Patented treatment
processes such as Vertech (section 3.6.3) capitalise on this effect even further.
A cautionary note needs to be sounded about the effect of oxygen injection on
high sulphur content wastewater in pipes. The addition of oxygen can lead to
conversion of sulphides and reduced sulphur forms to acidic sulphate
solution, which can cause severe corrosion of pipes (even, surprisingly, those
made of concrete).
Method 2 is very efficient in terms of gas consumption, achieving high levels
of dissolved oxygen efficiently but at some capital cost. It is becoming more
popular, being promoted by several of the larger gas companies. The addition
of oxygen to a sidestream allows conditions to be carefully optimised to
produce more or less full dissolution, whilst a helpful mixing effect is achieved
on pumping the oxygenated water stream back into the tank. The use of a
pressurised side-stream allows efficient oxygen injection with shallow tanks;
something that is difficult with method 1, as oxygen is lost to the atmosphere if
high levels of dissolved oxygen are attempted.
Method 1 typically involves some loss of oxygen gas but is the most common
method because it is low on capital cost and because it can almost always be
retrofitted to existing sewage plants. An ingenious touch in method 1 is to use
slots in a flexible plastic hose for the gas injection; when gas pressure is applied
these open up and allow gas into the sewage but on release of pressure the tiny
slots seal up again. This gives greater freedom from clogging; small bored
holes in rigid pipes tend to block easily when used intermittently. A typical
installation involves simply laying dozens of perforated plastic hoses across a
tank. The hoses are connected to a manifold pipe on the side of the tank.
All dissolution methods, but especially method 1, have limitations relating
to the behaviour of bubbles. Anyone who has visited a major science museum
recently may well have seen an educational toy in which bubbles of air are sent
up a large column of a viscous clear liquid. The equipment has viscous flow and
demonstrates that large bubbles rise faster than small ones according to
Stokes law for drag on a spherical object. However, even more obvious than
this is that the bubbles have a tendency to combine together after formation
but the converse, bubbles splitting up, never happens. Bubble coalescence
happens because the excess free energy due to the surface tension possessed
by two bubbles is less than that of one bubble of equivalent volume. Taking
the case of two identical spherical bubbles of volume V (radius R) and
overpressure coalescing, for example, the additional gas pressure energy
shrinks by 20%
where S is the surface tension and is the pressure inside the bubble above
the local hydrostatic pressure.
On coalescence, one bubble, pressure , radius and volume is
formed and
Hence
This effect limits the maximum flowrate through diffusers: if the flowrate
is too high, then bubbles appear too close together and they begin to
coalesce. The large coalesced bubbles rise more quickly, tend to coalesce
with other smaller, slow bubbles as they rise and coalesce further until the
neat regular streams of small bubbles seen at lower flowrates, which dissolve
very efficiently, are completely disrupted by irregular bubbles with large
gouts of bubbles rising to the surface incompletely dissolved.
With all these methods, a dissolved oxygen electrode suspended in the
tank or pipe being treated can feedback achieved oxygen level to a
controlling valve, thus economising on oxygen usage. Typically oxygen is
only required only for a few hours a day to handle peaks in BOD. However,
oxygen electrodes are subject to fouling and other failures in services in
water treatment and some care is necessary in choosing and locating the
electrode. The type chosen will typically be isolated from the water stream
by a plastic membrane that allows only small molecules, such as oxygen, to
pass, as described in section 2.1.7.
For continuous supply of oxygen to wastewater treatment, bulk liquid
pure oxygen supplies are regarded as rather expensive. More popular is to
use ~ 90% pure oxygen produced by pressure swing absorption (PSA).
These units can be made to operate with long intervals between mainten-
ance and are steadily growing in energy efficiency. PSA oxygen injection is
popular, for example, for oxygen treatment of tanks where there is a foul
odour problem. A conventional tank aeration system needs a similar power
for its air compressor or other agitation system to that needed for the PSA
oxygen system. However, with conventional aeration, nitrogen bubbling
out of the surface of the tank carries an aerosol into the air, efficiently
vaporising any foul-smelling molecules from the wastewater. Other aeration
systems which pump a fountain of water through the air have a similar
problem. With oxygen injection, bubbling at the surface can be reduced
virtually to zero, minimising odour problems.
Pure oxygen for peak demand. Even with gas from an efficient PSA, the
cost of continuous oxygen supply is still generally regarded as too high in
most cases and aerobic sewage treatment using air is the norm. Pure oxygen
treatment is becoming very common, however, for peak shaving. The
volume and BOD per of wastewater in many areas varies between
summer and winter in addition to shorter period variations. Although some
very short period variations are averaged out anyway by the natural water-
course, and by lagoons and settling tanks at the water treatment plant, long
period variations are more problematical. One attractive alternative is to
meet the base-load oxygen demand of the sewage stream with air, while
during peaks of high BOD or high flow oxygen is injected. This approach
optimises the compromise between running and capital costs; although the
oxygen is expensive, it is used only occasionally, whilst the main waste-
water plant can be sized only a little above average demand, rather than
having to be sized to meet peak demand. A typical liquid oxygen instal-
lation of this type uses an annual average of only a few tonnes per week of
oxygen but is capable of high rates of addition when necessary.
In many peak-shaving plants, oxygen injection is necessary only in
summer. In summer, water volumes are reduced so that even where there
is a similar absolute BOD there will be a higher BOD per unit volume.
This is just when the lower solubility of oxygen at higher temperatures
makes it more difficult to achieve adequate oxygenation. Furthermore, in
summer aerobic processes in water are faster and therefore have a higher
oxygen demand. Such considerations, along with rainfall and water-usage
patterns, determine the peak and average oxygen demands. The cost re-
duction achievable with an LO peak system depends on this peak/average
demand ratio but can be a factor of two or more.
Oxygenation of rivers and lakes. The River Thames in London, UK,
flows through hundreds of towns and industrial areas, and receives run-off
water from intensively farmed crop-growing areas. Despite this, the river
has a large population of fish, including even the occasional salmon. This is
only possible because oxygen is injected in tonnage quantities whenever
the BOD of the river rises too high. Just as oxygen injection can be used to
increased oxygenation in sewage works, it can also be employed to provide
respiratory oxygen to natural river microbes and fish.
In lakes with a relatively small throughput of water, small amounts of
organic contaminants, from sewage, food processing or paper-making, can
have serious consequences. These cause turbidity and excess BOD in the
water. Turbidity inhibits photosynthesis of oxygen. It is possible for BOD
to go up, dissolved oxygen levels to fall and living organisms in the lake to
die. This creates a whole new mass of organic material with its own BOD,
further exacerbating the problem. This classic vicious circle quickly results
in a dead lake. Conversely, if BOD falls to below the critical level, plants
and animals can recolonise the lake, resulting in the consumption of the
organic chemicals and the production of further oxygen, further reducing
BOD. Inland lakes with small flows are thus bistable. Small lakes can be
revived by tipping them out of the more or less dead anaerobic state into
the living aerobic state by a steady injection of oxygen over some weeks or
months.
3.6.2 Drinking water treatment
Water from most sources other than rainwater or direct from springs, clean
lakes and reservoirs needs some treatment before it can be guaranteed to be
healthy for drinking. Algae, bacteria and other suspended matter are
removed to a large extent by filtration, often in simple sand bed filters.
Addition of alum or similar salts in part per million quantities helps in the
coagulation of suspended matter and aids filtration. However, the water may
still contain some pathogenic bacteria. These can be killed by adding a very
small amount of chlorine. Increasingly today, however, consumers find the
taste of even a small amount of residual chlorine unacceptable. Also, there
are worries about the formation (in almost undetectably small quantities) of
potentially carcinogenic organochlorine compounds. Trihalomethanes are
one example of this. Chlorine dioxide and chloramine treatment would also
appear to reduce the production of organochlorine compounds, although as
these treatments both use the element chlorine there are still some worries.
In the future ozone will probably be the most important alternative to
chlorine in the killing of pathogens in water. Ozone water treatment is already
the largest commercial application for ozone. Oxygen or air is passedthrough
a silent high-voltage discharge, producing a gas mixture with a small
percentage of ozone. When this is bubbled through water in small amounts it
easily and quickly kills residual bacteria.
Ozone has the fundamental advantage that it is unstable and will always
have disappeared completely before delivery from the tap. In any case
catalytic reactors are available which complete ozone decay very rapidly.
Ozone-treated water is therefore completely pure, although ironically this
means that to retain bacteriocidal action along pipelines an amount of
chlorine (a small fraction of the normal amount) must typically be added to
ozone-treated water.
Ozone also has the advantage of being a more powerful oxidation agent
than chlorine. This means that any residual organic matter, that not oxidised
by the microbial treatment of sewage, is decolorised and deodorised more
completely than with chlorine. With ozone there are also fewer handling risks
relative to chlorine. Ozone treatment is, however, a more expensive and
more complex technology than chlorine treatment.
Hardening/stabilisation of drinking water. In many parts of the world,
depending on rock formations in the rain catchment area, water is naturally
hard, containing soluble calcium and magnesium compounds. Other areas
have the opposite problem: an acidic water without any calcium or
magnesium to buffer and prevent the pH from falling (much of Scandinavia
and Scotland, for example). Paradoxically, although hardness is generally
regarded as a nuisance, a small amount of hardness is actually beneficial in
that it suppresses pipe corrosion and reduces pick-up of potentially harmful
pipe and tank materials such as copper, lead and zinc. New European water
standards now insist on a certain low minimum of hardness so plants in
northern England, Scotland and Scandinavia will be adding lime
to attain hardness targets. However, lime is minimally soluble and practical
systems in fact add a lime slurry and inject producing one of the
principal constituents of naturally hard water, calcium bicarbonate (Ca(H-
which remains in solution. Typical treatment levels are of the order
of Good mixing is essential to ensure that is not lost before it
has time to react.
Stabilisation of lime-softened water. Although ion exchange resins are used
on a relatively small scale for softening water, this would be impossible on
economic grounds for municipal water supplies. They are, however, used to
treat water for hardness economically by using cruder processes such as lime
softening, discovered in 1766 by Cavendish. Lime softening is achieved by
adding CaO or its hydrate, the cheapest industrial alkali, to
waters containing magnesium and calcium bicarbonates. Calcitic or dolo-
mitic deposits are precipitated and the water is softened
However, these sparingly soluble compounds are not totally precipitated
and are carried on into the remainder of a water treatment works, where
they may cause severe problems by precipitating out in filters, pipes and
pumps. Stabilisation prevents this unwanted deposition by adding to
the water stream after a settling tank has accumulated nearly all the
particulate calcite and dolomite. The injected reacts with dissolved and
microparticulate carbonates and produces the much more soluble bicarbon-
ates,
58
although in much smaller quantities than the input water, and also
neutralises any excess lime that may have been used
If a modest excess of lime is always added, the addition can be
controlled by a downstream pH meter since the limed effluent will have a pH
of, say, 10 to 14, dependent on the amount of excess, and the can be
added to bring the pH back to, say, 6 to 9.
One wastewater which is used in probably the largest quantities
anywhere in industry is the cooling water used in power stations. The
difference here is that most power stations recycle most of their cooling
water. Addition of to acidulate the water slightly is common practice as
it helps to reduce scale, as in the lime-softening of water.
58
This reaction is the reason why limewater turns mi l ky with then turns clear again if the
stream is continued.
3.6.3 Industrial wastewater treatment
Some industrial wastewater treatment is biologically based, although an
intensification of the treatment process is often necessary. For example,
pure oxygen injection systems are sometimes seen in the treatment of
industrial wastewater with high BOD. The steady tightening of regulations
on the pollution of water in many countries means that occasional high BOD
discharges, which used to be acceptable, must now be avoided. Many
industrial concerns, breweries and abattoirs for example, with wastewater
problems are therefore now installing pure oxygen equipment. Oxygen
injection is typically used on an occasional peak-shaving basis with the
wastewater normally being allowed straight out to sewers.
Industrial wastewater treatment more commonly uses chemical rather
than biological methods. The high toxicity and wide variations of pH and
pollutant concentrations mean that conventional biological treatment is
difficult or impossible. Very often, however, industrial chemical treatment
of wastewater is used near the point of generation of the wastewater simply
to reduce toxicity and pH to reasonable bounds. The water can then be
safely and economically piped away for further treatment in the conven-
tional municipal sewage works. It is this chemical primary effluent treatment
in which gases have their most important role.
One technique for aerobic treatment of industrial wastes which may be
important in the future is the use of hydrostatic pressure. Without microbial
action the high hydrostatic pressure down a deep shaft can be used to
dissolve large amounts of oxygen in water. The Vertech process involves
pumping sewage down a shaft more than a kilometre deep, adding oxygen at
a few hundred metres down, and pumping it back up again. High pressures
and high temperatures are obtained without expensive pressure vessels by
natural hydrostatic pressure and the increase of temperature with depth.
This very efficient dissolution of oxygen at high temperatures allows an
increased rate of reaction. The first commercial-scale pilot plant in
Apeldoorn, Netherlands, is now operating, treating 100 tonnes per day of
5% sludge water with considerable oxygen demand. Similar non-biological
oxygen treatment using surface equipment is used here and there for
industrial wastewater but has never caught on on a large scale.
Ozone is also now used in industrial wastewater processing, in order to
achieve limits imposed by local authorities for discharge to sewers and water
courses. The ease with which ozone can convert potentially harmful organic
substances to and water, and even highly toxic ions such as cyanide to
the relatively harmless cyanates, makes it very attractive as a small-scale
treatment for industrial discharge.
3.6.4 Control of pH with
One environmentally friendly way of controlling the pH of wastewater
streams is to add Many waste streams have too high a pH, i.e. alkali
pollution problems. In addition, the pH varies widely, depending on the
stage of the process. For example, a cleaning process might typically involve
caustic soda followed by rinsing.
Carbon dioxide forms the weak acid, carbonic acid, when dissolved in
water. Chemical equilibrium principles mean that addition of a weak acid to
an alkali-contaminated stream will both reduce and stabilise the pH. Even
low-pH wastewater streams are sometimes treated with by predosing
with trace amounts of alkali, followed by precise control of pH with
addition. Carbon dioxide is convenient because it can be easily controlled, is
non-toxic, inexpensive and readily available. A simple pH control rig might
consist of little more than a control valve supplying gas to a bubbler in
the waste stream, with a pH electrode downstream giving a feedback signal
to the control valve.
Wherever weakly acidic salts, such as ammonium chloride or ammonium
sulphate, are used to control pH, may be a possible candidate as an
environmentally sounder substitute. For example, can be used instead
of ammonium chloride to reduce pH in leather treatment. After removing
hair and then pickling in alkali (liming), leather has to be gradually
neutralised (delimed) before tanning under acidic conditions with sulphuric
acid/sulphates followed by dichromate. can sometimes replace the
more traditional ammoniacal solutions by reacting with the and
lowering the pH from 12 to about 7, when the acidic treatment can begin.
The ammonium salts are normally discharged into water courses where they
may cause significant smell and pollution problems. avoids these
problems totally and is less expensive.
can also, in one application, replace chlorine. Chlorine is used in
swimming pools, in much the same way as it is used in drinking water.
However, here again, the taste of chlorine, and its effect on the sensitive eye
membranes, is undesirable. The level of chlorine is normally maintained by
injection of small amounts of chlorine at intervals, the level being
maintained to control the pH (acidity) of the pool water. A more recent
scheme has been devised whereby the pool pH can be controlled by means of
with bacteriostatic properties being maintained by addition of liquid
and solid chemicals. This relatively new scheme means that gas cylinders of
chlorine need not be handled and the small risks of excessive chlorine leaks
or excessive addition are avoided.
3.6.5 Gas stripping
As mentioned above, certain industrial uses of water require processing
beyond the achievement of dri nki ng water standards. The removal of
dissolved gases, particularly oxygen, is one example of this. The removal of
volatile materials and dissolved gases from water can sometimes be achieved
by the elementary expedient of bubbling an inert gas through the water. Gas
stripping or sparging is used especially for aqueous product streams
(section 3.4.3). Often air is sufficiently inert that it can be used in gas
stripping; unless, of course, the gas to be stripped is oxygen.
Gas stripping is most effectively carried out in counter-current packed
columns but the capital cost of this kind of installation often means that less
efficient simple bubbling tanks are used. An alternative method is to add
inert gas to the low pipework somewhere in an installation and then allow
the sparging gas to escape at the top of the system, perhaps from the
headspace of a tank.
An adjustment of pH upwards will enable dissolved alkalis, such as
ammonia, to be stripped out by nitrogen sparging, whilst addition of acid to
lower pH will enable removal of dissolved volatile acids, such as hydrogen
sulphide or cyanide. The use of nitrogen for the stripping operation is
justified where oxidation, often of other contents of the wastewater, must be
avoided.
Gas stripping is seen in semiconductor water purification, in the
manufacture of soft drinks and in the preparation of boiler feedwater. In
semiconductors (section 3.5) absolute purity is important, whilst in boilers
the use of water which will not cause oxidation/corrosion of the boiler tubes
is very important. In semiconductor and boiler applications an alternative to
sparging is the addition of a reducing chemical; perhaps after an initial
gas-stripping operation. A typical chemical dosing is with hydrate of
hydrazine, the reaction producing simply nitrogen and water with dissolved
oxygen. Clearly, however, sparging is superior in that it does not involve a
hazardous chemical and any carry over of the sparging gas will be harmless,
whereas hydrazine carryover could be highly undesirable.
3.6.6 Oxidation treatment
Even with pure oxygen instead of air, oxidation treatment of industrial
effluent can be a slow reaction. The exception is the oxidation of iron and
manganese in which oxidation followed by precipitation is a possible
treatment. Interestingly, iron and manganese ions can be deliberately added
to organic-contaminated streams where they act as catalysts to allow the
oxidation of oxygenated organics, alcohols, aldehydes and ketones, which
otherwise react much too slowly. Similarly, transition metal catalysts can be
used to catalyse oxidation of sulphides in wastewater.
Ozone treatment. Except with biological treatment, and with the excep-
tions noted above, oxygen is often ineffective in industrial wastewater
treatment. It is insufficiently active as an oxidant and it is too difficult to
dissolve in water. If the oxygen is converted into ozone, however, oxidation
is almost invariably successful. However, ozone has all the problems seen in
drinking water treatment. It is expensive to make, whether direct from air or
from oxygen. Furthermore it is difficult to dissolve in water, though more
soluble than oxygen, and it rapidly decomposes back, in a matter of minutes,
to unreactive oxygen again.
An example of where ozone is used is in phenol and cyanide oxidation.
Ozone neatly converts many toxic and persistent phenol products, such as
occur in phenolic glue manufacture and use, into harmless and readily
biodegradable smaller molecules. Similarly, cyanides are converted to much
less toxic cyanates. Ozone system suppliers now claim reductions of serious
phenol pollution down to ppb levels with ozone injection.
Paper-making and wastewater. Another area in which ozone and oxygen
may make inroads in the future is in the paper-making industry. The
manufacture of paper is associated with the consumption and potential
pollution of very large quantities of water. With all paper-making processes
there would appear to be good opportunities for much greater use of
industrial gases. Large amounts of water end up containing suspended
organic matter and the waste stream has a very high BOD. This effluent can
be treated by conventional aerobic biological plant or with pure oxygen
biological plant
The basic paper-making process is the dissolution of mashed heartwood
from trees (the bark is discarded). Many variations on the process exist but
the following is representative. In Kraft pulping, wood chips are cooked
with aqueous sodium hydroxide/sodium sulphate under pressure at ~ 150C
for a few hours. The cooking process destroys the lignin, the resin that binds
the cellulose fibres of the wood together. The released fibres are more or less
pure cellulose fibres but they are mixed with chemical by-products: the alkali
that was added, sulphide formed from the sulphate, and evil-smelling
mercaptans and complex alcohols with a dark colour known as black
liquor. The processed pulp is washed and the black liquor is recycled by
evaporation followed by separation into cooking chemicals, waste lime
sludge and tall oil, the latter containing resins and fatty acids from the
wood, including turpentine. A more recent variation on Kraft pulping uses a
sulphite mix, blowing through the mix, after adding magnesium
hydroxide and/or lime.
After the initial chemical cooking process has extracted cellulose from the
wood-pulp, purified cellulose is usually bleached white with an oxidising
agent, commonly chlorine dioxide. In many cases, chlorine-based industrial
treatment processes are being frowned upon today because minute amounts
of potentially harmful chlorinated organic compounds can be formed.
Dioxins, for example, have been found (probably in harmless trace
amounts, it must be admitted) in the effluent from paper mills and in the
paper itself.
Oxygen can sometimes be used and the bleaching of lignin by oxygen/
caustic soda is a now a proven process, replacing either hypochlorite or
chlorine dioxide usage. Where oxygen is an insufficiently powerful oxidant,
ozone can often be effective. Ozone is an effective bleaching agent and can
be used both in paper fibre and textile fibre bleaching, replacing chlorine or
hypochlorite. Recently, babies disposable nappies came under the spotlight
because of fears of organochlorine contamination due to chlorine-based
bleaching. Ozone or oxygen would guarantee that organochlorine com-
pounds would not be present even at exceedingly low levels but would be
entirely absent.
3.7 Industrial cryogenics
3.7.1 Properties of liquefied gases
The properties of common cryogens are important parameters for the design
of cryogenic apparatus. There is in fact only a rather small set of acceptable
cryogens. Even in the short list given in Table 3.9, hydrogen and methane
are flammable, neon is rather expensive and is barely cryogenic (195 K
is, after all, only 78C).
There are a number of things which Table 3.9 emphasises. One is the wide
variation in latent heat of the materials, although the exceptionally low
density of helium and hydrogen mean that their latent heats are actually
lower per unit volume. Even after the addition of more exotic molecules to
this short table, it is noticeable that there are a number of gaps in the
spectrum of boiling points. There are in fact numerous slightly less ordinary
gases which can be used to fill these gaps from 77 K upwards. However, if a
controlled temperature bath is needed at 12 or 50 K, for example, then the
convenience of boiling cryogen is not available.
3.7.2 Cryogenic phenomena
Remarkable effects occur at temperatures just a few degrees below those
normal on earth. Below 0C a large number of things happen related to the
freezing of water. The vast bulk of university cryogenics research is devoted
to the fascinating, but mainly esoteric, pheonomena that occur at tempera-
tures below the 4 K boiling point of liquid helium. However, in industry
cryogenics is almost entirely confined to the region above the boiling point of
liquid nitrogen, i.e. above 77 K.
A number of important phenomena happen at industrial cryogenic
temperatures:
many ordinarily flexible or tough materials become brittle,

59
although
some non-ferrous metals, such as aluminium, become stronger at
cryogenic temperatures;
most gases, except the permanent

60
gases, either condense or freeze;
ceramic copper oxide materials become superconductors;
high rates of cooling from room temperature are possible because of the
very low temperature relative to room temperature (195C) and high
heat capacity of LN;
most materials shrink quite markedly in linear dimensions relative to

room temperature.
In a few research groups, such as that at Southampton University, UK,
phenomena at 77 K are studied in detail. Recent work in this area has been
directed towards topics such as superconductor cooling and enhanced
boiling surfaces. For example, special precautions are necessary during the
cool-down of the fragile ceramic materials because differential thermal
expansion can lead to cracking. Enhanced boiling surfaces, i.e. surfaces
capable of even higher rates of cooling than expected hitherto, may be useful
in enhancing current cryocooling equipment, as well as for future yet-to-be-
determined uses (section 4.3).
59
One of the less familiar of ice-related phenomena relates to the extraordinary strength
achieved by freezing quite dilute suspensions of fibrous materials in water. Although it is more
normal to utilise the weakening of materials entailed by LN freezing, it is worth remembering
that some materials increase in strength when deep frozen. Some frozen food, for example,
seems to be every bit as strong as wood. The British 1940 wartime Habbakuk project relied on
pykrete, a 3% wood-pulp in ice composite, which was shown to melt more slowly than ice,
have many times the strength and to resist explosive detonations to a large extent. The project
was intended to produce million tonne ships/icebergs to act as temporary invasion craft and
aircraft carriers.
60
Permanent gases are nitrogen, oxygen, argon, helium, hydrogen and neon.
Boiling. LN will boil vigorously when heated and the boiling process is
what allows the spectacular rates of cooling that are the principal reason for
using LN. However, boiling is, in detail, a subtle process. For example, the
bubbles that occur in heated water are not necessarily due to boiling.
Actually, the smallest of the bubbles, when they first appear, are dissolved
gas from the water being evolved. It is difficult to see exactly what is
happening when a saucepan is really boiling because of the violent
disturbance of the surface if the heat is at all strong. For boiling to occur,
molecules of a liquid have to retreat away from a cavity that is sufficiently
large that surface tension forces on the forming bubble cannot immediately
collapse it. The overpressure required to form a bubble is smaller for larger
bubbles. As the bubble gets bigger, it becomes more stable. The vapour
pressure of liquid inside a bubble depends on the bubble radius, tending to
the equilibrium vapour pressure at large radii and tending to zero as the
bubble shrinks. Small bubbles are unstable, tending to collapse back into
liquid.
Superheating is explained by reference to what happens as the vapour
pressure begins to exceed the ambient pressure. Bubbles, once formed, will
grow if above a certain critical size. However, if nothing happens to form
these microbubbles, then nothing happens at all except the liquid becomes
hotter, i.e. the liquid is superheated. The addition of a surface on which the
minimum size bubbles can be formed with lower energy than in the free
liquid causes nucleation, i.e. a stream of bubbles originating at particular
points on that surface. Nucleation around the rows of ions left behind by a
high-energy ionising particle is the basis for the cryogenic bubble chamber.
Generally filled with liquid hydrogen, these were used in many of the 1960s
and 1970s fundamental particle physics discoveries on which current
quark/gluon theory is based. Particle collisions can be tracked by arranging
for flash photographs to be taken of bubbles in the superheated liquid
hydrogen just after a collision occurs inside it but before general boiling
occurs (Figure 1.6).
Boiling proceeds in a number of different regimes depending on the heat
flux applied to a liquid: convection, nucleate boiling and film boiling. At low
heat flux inputs boiling is indistinguishable from normal gentle heating.
Heat is applied at the bottom of a bath, convection occurs, carrying the
heated liquid from the bottom to the top surface, where evaporation takes
place enhanced by the increased temperature.
The nucleate boiling regime exists because of the very high overpressure
required to cause a very small bubble of vapour to exist against surface
tension. If a nucleation surface exists, this lowers the overpressure required
so boiling occurs selectively on surfaces, around points on surfaces or on
suspended particles. To see precisely why this is so, consider the equilibrium
of forces on a bubble cut in half once it has formed
Figure 3.32 Typical experimental pool boiling curve for nitrogen (Hands, 1986).
Force on half a spherical bubble due to excess vapour pressure inside
Force on half bubble due to surface tension, S, per unit length
where R is the diameter of the bubble. At equilibrium the excess pressure
required to form a bubble is simply so for very tiny bubbles a
large overpressure is needed, implying a necessity for strong superheating.
If an overpressure of 0.1 atm is assumed to be needed, then for LN, despite
its low surface tension the minimum bubble size is already
0.6 mm. However, if a surface can prefabricate or nucleate a microbubble
of smaller mi ni mum radius, then boiling can occur closer to the theoretical
boiling point.
Enhanced boiling surfaces seek to achieve exactly this, in the most efficient
way possible, by applying (typically) a porous surface, which is easily wetted,
will not easily show the Leidenfrost phenomenon and provides the maximum
nucleation effect. Enhanced boiling surfaces have applications in improving
the performance of heat exchangers used with LN in cryogenic cooling,
perhaps in improving the performance of plates and packing in distillation
columns, and perhaps in new devices such as superconductor fault
current limiters, which seek to control cryogenic temperatures very
accurately by boiling liquids (section 4.3). One way of making enhanced
boiling surfaces is to create a spongy metal surface on a metal.
At very high heat fluxes the heated surface no longer contacts the liquid
directly at all. A thin film of vapour forms which isolates the surface and
partially insulates it. This is, in effect, the Leidenfrost phenomenon. The
transition to film boiling, as it is known, is marked by a sharp decrease in
heat transfer (Figure 3.32).
Vapour pressure and boiling point. As with all liquids, cryogen vapour
pressures increase with temperature approximately exponentially (section
1.3.14). This means that their boiling points depend on pressure. For
example, for liquid nitrogen boiling point varies with pressure as shown in
Table 3.10.
It is possible to control chemical reaction temperature to a fraction of a
degree using pressure control of a boiling liquid at elevated temperature (for
instance, in organic chemical reactions). Subatmospheric pressures can also
be used to control boiling point. By vacuum pumping, LN can reach 65 K,
instead of 77 K, whilst LHe can reach 1 K, rather than 4.2 K. Controlling
pressure above a cryogen bath via regulators and pumps, whilst maintaining
boiling by means of enhanced boiling surfaces, is a possible means of
accurate temperature control.
Flash loss. The change of boiling point with pressure is also important
because of the loss of LN that occurs when LN from pressurised tanks is
decanted into open vessels. The resultant losses are known as flash loss
because violent flash boiling occurs when the liquid is first decanted into
another vessel at atmospheric pressure.
A liquid cryogen is generally stored at a high pressure; liquid nitrogen is
frequently stored at 15 barg or more, for example. This may well be
convenient for supplying gaseous nitrogen to a process at that kind of
pressure. Storage vessels are almost always well above atmospheric pressure
because handling problems are eased in any case; the gas pressure in the
ullage space can be used to drive liquid transfer rather than having to rely on
cryogenic pumps.
However, when a pressurised tank is needed to supply liquid nitrogen at
atmospheric pressure, there will be a considerable amount of boiling when
pressure is released from the liquid stream, and a sizeable fraction of the
nitrogen extracted will be lost, until the remaining liquid has cooled to the
boiling point at 1 bar pressure.
Figure 3.33 Flash losses in liquid nitorgen decant from pressurised tanks.
The flash loss can be theoretically calculated and these figures are often
seen tabulated (Figure 3.33). However, the theory is only accurate if fairly
large amounts of LN are withdrawn, heat-leaks in the LN decant pipework
are small, the receiving vessel is pre-cooled to cryogenic temperatures and
precautions are taken to ensure that droplets of LN are caught and drained
back into the receiving vessel. The conditions under which cryogens are
normally withdrawn from highly pressurised vessels rarely meet these
criteria. In general, therefore, much larger amounts of cryogen are wasted
than Table 3.11 indicates.
A caution should also be sounded about the accuracy of the figures in
Table 3.11. Several major data sources consulted for data disagree with one
another by 5 or 10%. The data in Figure 3.33 differ slightly from those in
Table 3.11, for example. However, given the other uncertainties, greater
accuracy is probably unnecessary. The lesson should be that, whatever the
losses, they are surprisingly high. Decant systems often try to restrain
losses by using insulation and some sort of counter-current system for
cooling incoming warm liquid with boiling LN vapour.
Flash losses in Carbon dioxide sublimes at 195 K at 1 bara pressure.
The cryogenic use of carbon dioxide is to some extent limited by its lack of a
liquid form at atmospheric pressure. A considerable pressure on the gas
(5060 bar) is required to retain it in liquid form (section 2.4.7). The property
of sublimation is useful, though, in small applications such as keeping small
amounts of food cool simply and without mess. No mechanical refrigerating
device is yet available which will work without an external power supply in the
size range from a fraction of a kilogram to a few tens of kilograms.
Alternatives that do not involve sublimation are invariably more complex to
apply. Blocks of ice have the snag of needing some way of handling the water
produced, whilst cryogenic liquids such as LN cannot simply be installed in a
small block in an insulated product package.
Liquid has considerably less cooling capacity than solid For
example, liquid stored under its own vapour pressure at room
temperature will suffer a flash loss of 70% (theoretical minimum) by mass on
release of pressure. Even from a refrigerated storage tank at 10 bar or so, a
large flash loss is expected. A relatively large volume of liquid is therefore
needed to create a fairly small block of dry ice. This is not immediately
apparent as the dry ice snow produced by simple expansion in a snow horn
has a very low density.
A snow horn is used to produce snow in situ on a material or device to
be cooled. Many applications employ liquid because of the ease of
handling the liquid but use a snow horn to apply it. The snow horn is an
expansion nozzle with a shield to prevent too much splash back from the
nozzle. The nozzle itself is arranged so that there are no constrictions that
could get blocked with solid and is sized to restrict flow to a suitable
value. The snow produced has a fluffy texture and is of low density,
always less than 50% of theoretical density.
Small-scale applications can use the compressed form of solid (dry
ice) which is much more compact than . Solid of near-theoretical
density (around can be produced by compressing snow in a
hydraulically-driven piston and cylinder arrangement. The resultant large
block of dry ice is then ejected from the cylinder and sawn up into slabs 25 or
50 mm thick, which are then individually wrapped in paper so that they will
not weld together in transit. Solid in slabs is, as a result, rather
expensive but does achieve the maximum possible refrigeration capacity in
the minimum space. Less laborious is the production of dry ice in pellets.
These are also produced by compression of dry ice snow but the pellets,
typically a few millimetres in dimension, are simply ejected directly into a
bag inside an insulated cardboard box for delivery. The gas space between
the pellets means that this form is not as space-efficient as solid slabs of
Small-size pellets are used in dry ice bead blasting (section 3.7.4).
Embrittlement. Many tough plastics and some tough metals become very
brittle below room temperatures. Plastics generally have a glass transition
temperature below which they are glassy rather than tough or rubbery. The
glass transition temperature reflects the fact that atoms in the polymer
molecules are no longer free to slide over each other (see Figure 2.67).
Ordinary carbon steels with a high degree of toughness at room temperature
are severely embrittled at low temperatures, whilst cast irons and similar
materials are rendered more brittle by cryogenic immersion.
It is also worth remarking that differential thermal contraction caused by
sudden immersion in cryogens (thermal shock) is a frequent cause of
materials failing. As noted above, many normally tough materials become
less able to withstand this kind of internal stress because of brittleness. At
the same time, the thermal conductivity of materials decreases rapidly as
temperature falls so that relief of stress by conduction is slow.
The modification of many plastic materials by the addition of large
amounts of fillers, such as sand or glass fibre, is an important technique for
rendering plastics more useful at cryogenic temperatures. These fillers often
increase the thermal conductivity and always decrease the thermal expansi-
vity of the plastic, decreasing the effects of thermal shock. Typical values of
2% contraction from room temperature to 77 K can be reduced by a factor of
10 or so by adding large percentages of filler.
Turbulent flow. As noted in section 1.3.7, many cryogens have a rather low
viscosity. LN is one seventh of the viscosity of room temperature water, for
example, whilst liquid helium is 300 times less viscous than water, although
in this case the low density of LHe means that the viscous forces still have
some importance. These extremely low viscosities mean that cryogens only
behave as viscous liquids at low speeds and it is very common to have
turbulent flow of cryogens in pipes and vessels, which has a number of
consequences. Not only does this lead to a requirement for larger than
expected pumps and pipe sizes but it may mean that some allowance will be
needed for cryogen losses due to the additional heat input caused by the
turbulence.
Vacuum-insulated pipework and phase separators. In order to deliver
liquid nitrogen or other cryogens to the point of use without unnecessary
heat loss, it is fairly obvious that vacuum-insulated pipework is advan-
tageous. It is perhaps less obvious that vacuum-insulated pipework may
actually be essential. If an uninsulated pipe is used for a steady flow, the
amount of boiling along its length may well be such as to cause very unsteady
turbulent flow with plug and churning flow. Such an unsteady two-phase
flow is difficult to measure and control effectively. Furthermore, if an
uninsulated pipe is used for unsteady demand, then control problems may
be insurmountable. Every time the pipe flow is stopped, it will boil dry and
begin to warm up. After a few minutes, if flow is demanded, liquid will flow
into a warm pipe and instantly vaporise so huge amounts of gas will pass
down the line. The blast of gas cools steadily until droplets of liquid cryogen
appear in it. It will often be several minutes before liquid cryogen flows out
of the end. These great blasts of gas are often totally unacceptable in
operation. In an operation such as can rigidifying, for example, a precision-
measured few drops of liquid cryogen must be guaranteed every time the
dispense valve opens and therefore any gas flow is unacceptable.
Vacuum-insulated pipework is made in a similar way to vacuum-insulated
vessels. It has inner and outer pipes of (typically) stainless steel, the void
being evacuated, sometimes with a powder or metal-foil fill. Long sections
are welded up and then evacuated, either in the suppliers facilities or in situ.
At the end of a run of any length a bellows section is normally included on
the inner pipe, allowing the inner pipe to contract relative to the outer pipe
when the cryogen is admitted. Sometimes as with cryogenic tanks, points are
left where a vacuum pump can be reattached to restore vacuum quality.
Vacuum-insulated pipework is more difficult to install than ordinary
pipework. Tees and elbows are often executed in ordinary pipework with
plastic foam insulation, the vacuum insulation resuming wherever there is a
straight run. Fairly gentle bends can be fabricated by filling both pipes with
water and freezing, prior to bending, or by using double pipe with a polymer
string spiral wrapping around the inner pipe as a quasi-continuous space.
Demountable joints are possible in vacuum-insulated pipe. A demountable
joint in an ordinary pipe costs only a few dollars. In vacuum-insulated pipes,
however, these are complex assemblies costing many hundreds of dollars.
Installations are only effectively carried out by a few specialist firms and
major gas suppliers.
Even with vacuum-insulated pipework there is still typically a certain
amount of two-phase flow, not pure liquid flow, at the point of use. Many
installations will benefit from a phase separator near the point of use. This is
often no more than a small insulated tank with liquid inlet and outlet and a
vapour vent, although designs with plates and baffles can behave better in
bumpy flow.
3.7.3 Cryogenic practice
Environmental cabinets. It is frequently the case in manufacturing that a
system or component must be cooled down or heated up and tested to check
that it will function when the ambient temperature is low. An insulated
cabinet, fitted with a circulating fan and powerful heater, serves for the
higher temperatures but lower temperatures are less easily achieved with a
simple refrigerator set-up. Either the refrigerator will take a long time to
cool or it will be difficult to regulate the temperature precisely. An LN-fed
cabinet can be cooled very swiftly indeed and controlled accurately. In the
electronics industry it is often a requirement for military circuits to be tested
from 55C to 125C and it is important to be able to carry out this testing
APPLICATIONS OF GASES IN INDUSTRY 451 1
swiftly. LN is often used in small environmental cabinets for this purpose.
Thermal cycling tests, with hundreds or thousands or more of cycles from
cold to hot temperatures, are often specified during the development of
electronics. A small cabinet with LN cooling and powerful electric heating is
typically used. For small cabinets liquid is an alternative that is most
suited to cabinets which are used only intermittently.
LN also has the advantage of a low capital cost for large cabinets, which is
often appropriate as large-scale cold test facilities are only occasionally
required in most companies. The largest LN cabinets are big enough to
contain a medium-sized aircraft. Testing of aircraft in different ambients is
very important as they encounter very low temperatures in the upper
atmosphere, even above the equator. Typically, aircraft testing has a highly
intermittent use pattern which makes an LN system very appropriate.
Related to LN environmental cabinets for aircraft testing are the small
supersonic wind-tunnel test and experimental facilities run by universities
and institutes. An inexpensive way to construct a supersonic tunnel is to use
a large capacity vaporiser and a high-pressure pump, if necessary, to achieve
the bars or tens of bars pressure required for supersonic flow. Wind tunnels a
few centimetres across are possible using LN so that even the smallest
research laboratories and universities can now afford supersonic facilities
(see section 4.3.2).
LN vacuum cold traps. The use of a cryogen to so cool a gas that it
condenses and creates a vacuum is a very obvious idea that has been used, at
least in research, ever since cryogens became available. The cold trap is the
simplest embodiment of the principle.
Cold traps are still a very low-cost means of pumping away unwanted
vapours out of a high-vacuum chamber. Installed before a high-vacuum
pump, the trap consists of a liquid nitrogen reservoir thermally linked to a
set of metal baffles. It prevents backstreaming of oil vapour from the pump
to the chamber and condenses vapours from the chamber onto its baffles. In
this way, the pump needs to extract only permanent gases and can be much
smaller than would otherwise be the case in a vacuum system. Periodically,
when high vacuum can be interrupted, the trap is warmed to room
temperature, the absorbed vapours evaporate and are pumped away easily
at relatively high pressures. Depending on the pumping load, it is possible to
accumulate a dangerous mix of gases, which could explode or react
dangerously when warmed up for regeneration. This would happen, for
example, if the pumping load were air, followed by slow outgassing of
hydrocarbons. Regeneration cycles, therefore, should be carried out as
frequently as is reasonably convenient and precautions, such as a flow of
nitrogen purge gas, taken if necessary.
Cryopumps. Proper cryopumps are more sophisticated in two respects
compared to simple cold traps. First, they employ not just metal baffles but
also high surface area absorbers. Second, they normally use lower tempera-
tures than the 77 K of liquid nitrogen, either by using liquid helium or by
means of a Stirling cycle refrigerator.
Some ingenuity is employed in the design of cryopumps to ensure that the
heat load on the refrigerator is almost entirely that due to the condensation
of gases. To achieve this, radiation shields (optical baffles) are placed
between the cold surface and the (warm) volume being pumped. These
radiation shields are cooled to intermediate temperatures by LN or by an
additional Stirling stage (see Figure 3.34). In many pumping jobs the gas
load is dominated by water vapour. This is condensed on the first
intermediate temperature baffles, reducing the heat load on the cold panels
and thus increasing the thermodynamic efficiency of the system. The
absorbers are generally coated onto the cold absorber panels and are usually
made from some kind of microporous carbon. Further advances are now
being made in the thermodynamic efficiency of cryopumps, with many now
using the GiffordMcMahon cycle for the helium refrigerator.
The cryopump is capable of pumping millions of litres per second and as
such is probably the most powerful high-vacuum pump available, provided
the connecting pipe bores and paths are such as to allow it to display its
potential. Large vacuum chambers, such as outer space simulators for
spacecraft testing, for example, use cryopumps. The power of the cryopump
concept is illustrated by the looking at the vapour pressure of gases. At
liquid helium temperatures only helium and hydrogen have any vapour
pressure that is measurable and LH has only mbars.
Stirling cycle cryorefrigerators. The liquid helium temperatures needed for
cryopumps are easily achieved with merchant-delivered liquid helium and
this is ideal for intermittently used apparatus. However, for more or less
continuous vacuum pumping a Stirling cycle engine using helium as the
working fluid is the appropriate solution (Figure 3.35). These Stirling
engines are also used as a more or less consumable-free variable tempera-
ture cryostat at all temperatures below 77 K, down to 20 K at reasonable
heat loads (a few watts or so) and to a few degrees above helium boiling
point (4.2 K). To achieve these low temperatures 2 or 3 stages are needed,
with the colder stage rejecting heat to the hotter stage. Although their
efficiency is generally low and they are mechanical devices, Stirling cycle
cryorefrigerators can be made quite reliable and they are very popular in
scientific research where heat loads are very small.
A Stirling engine works by compressing the working gas isothermally,
rejecting heat to the hot side, then forcing this gas at constant volume
through a regenerator, a simple macroporous metal mass, often fine wire
windings or similar, which cools the gas. The cool compressed gas is then
allowed to absorb heat from the cold side and is thereby expanded
Figure 3.34 (a) Cut-away view of a cryopump: 1, inlet flange; 2, inlet louvre (80 K); 3, second
stage cryopanel leaflets (15 K); 4, hydrogen vapour pressure thermometer; 5, purge accessory
port; 6, first stage cryopanel (80 K); 7, vacuum shroud (courtesy of Edwards High Vacuum), (b)
Two cryopumps (left, with helium compressors bottom left) evacuate a large process chamber.
Note the gate valves for isolation above upper pump and below lower pump (courtesy of Oxford
Plasma Technology).
Figure 3.35 Stirling cycle principle.
isothermally. The expanded gas is then forced at constant volume back
through the regenerator, cooling this back down and restoring the initial
starting position ready for another cycle. A simple implementation of the
principle can involve the use of a moving piston and a moving regenerator
operated by the same drive shaft and a very simple mechanical arrangement.
Pipe-freezing. This is an important maintenance procedure useful for
many pipework systems, from the tiniest 6 mm line to oil lines 1 m or more in
diameter. Pipe-freezing enables modifications to be made to pipelines
without shutdown or draining off the system. First, the system must be full of
process liquid or water. The liquid flow is then stopped and liquid nitrogen
poured into a box surrounding the pipeline. Within a few minutes or hours
(depending on the pipeline size), a plug of frozen liquid forms and is
typically sufficient to resist a minimum of a few bars overpressure. On long
freezing times the freezing box must be topped up with LN from time to
time. Once frozen, the LN in the box will cease to boil so vigorously, the
pipeline can be broken into and a small amount of process liquid drained off.
Necessary operations, replacing valves, welding on new pipeline connec-
tions and so forth, can then be carried out. The LN box is then drained and
removed. The frozen slug melts, freeing flow and the system can once again
be operated as normal.
There are a few potential pitfalls for the unwary, however. Although a
6 mm line containing water can be frozen and blocked in under a minute, a
large 300 mm line may take some hours because there is roughly a square law
relation between freeze time and pipe diameter, as expected from theoreti-
cal calculations from simple assumptions. This can be seen by assuming the
growth of a cylindrical plug of length L in a radial direction, with outer pipe
wall at and inner bore at 273 and integrating to
find the heat flux Q' through a thick-walled pipe. Further integration gives
the time for the ice to fill the pipe bore. The first integration yields
where K is the thermal conductivity of ice. On integration to measure how
long it takes an ice plug to form in a water pipe at 0C, the following
result is obtained
where is the ice density and s is its specific heat of fusion. The cold
temperature might be taken to be 77 K, and to be 273 K for an
illustrative calculation, giving 2 h for a 60 mm inner diameter pipe. This
formula, and ones like it, tend to underestimate freezing times. More
accurate calculations take account of residual flow, end effects (only in the
middle of a long bath would the above strictly apply), pipe wall conductivity
and convection effects, all of which make freezing somewhat slower.
Prudence dictates that a further safety margin of a factor of 10 or more might
be wise where possible or a trial conducted first.
Tables giving freeze time are available in a limited number of cases. The
line pressure that the frozen plug can withstand increases as the plug length
increases. Rules of thumb are used in which nominal pressures of 8 bar (in
oil, which is more slippery and does not expand on freezing) or 16 bar (in
water) can be withstood by a plug only two diameters long. At higher
pressures, it would be prudent to allow longer plugs on a pro rata basis.
Freezing data assumes absolutely zero flow but often flow will be present;
perhaps because a leaking valve is the reason for the need for the
pipe-freezing for repair.
Table 3.12 provides a guide to the time required for freezing water in
pipes.
It is possible to damage a pipeline by freezing. Steel pipelines are
embrittled and can be shattered with a careless blow with a hammer or
excessive torque from a spanner. Some plastic pipelines can be frozen but
some will shatter. Ice expansion is not normally a problem, provided the ice
plug has room to expand sideways. Forming an iceplug too close (less than
10 diameters) to a closed valve may cause a pipe to split as the ice is then
confined sideways. Finally, once the LN box is in place to freeze the pipe,
access to the trench in which the pipe is situated should be restricted, as the
cold evaporated nitrogen gas will lie in the trench and could cause a fatal
accident by asphyxia. Plenty of time should be allowed for the gas to clear
before carrying out work on calm days or when working indoors.
The group at the University of Southampton, UK, amongst others, has
carried out both theoretical modelling and practical experiments on
pipe-freezing with a variety of pipeline substances and pipe sizes, including
situations where there is some residual flow.
Ground-freezing. Occasionally, it is required to tunnel in soil or rock of
very low strength. Alternatively, a deep hole or trench is needed and there is
not room to excavate a large hole with gently sloping sides. There are a
number of alternatives in this situation. However, one straightforward
alternative is to insert small bore pipes into the soil to be tunnelled or dug
and inject liquid nitrogen for hours or days. Eventually, the water content of
the ground is fully frozen and binds together the loose soil and rock so that
fairly conventional tunnelling, with little danger of roof or wall collapse, can
be carried out. The process is quicker than using brine, although typically
more explosive.
Cryogenic shrink-fitting. Many engineering assemblies rely on an inter-
ference fit. Briefly, this means that the surfaces of two components
nominally overlap by a few microns when those components are assembled,
i.e. an axle of 10 mm (10 000 m) diameter will have an interference fit of
with a wheel with a hole of diameter that is 10 mm thick. An
interference fit is often achieved by force-fitting, i.e. using a hydraulic press
of many tonnes to force the components together. However, force-fitting is
more likely to result in scrap. The standard hot process for shrink-fitting is to
heat the housing, whilst keeping the insert cool. Once the housing is hot, the
insert is pushed into place and the whole assembly is allowed to cool until the
insert is held in place by the shrinking of the housing around it.
Cryogenic shrink-fitting is a gentler alternative to the standard process
and one which is particularly economical for inserting a small component
into a larger housing. Rather than warming up the large housing, the small
component is immersed in LN and allowed to shrink. The size of the smaller
component is calculated so that when it expands to normal size it will fit
immoveably in the larger housing.
An example is easily calculated to show the efficacy of the method. If a
10mm shaft made of brass is cooled to LN temperatures, it will shrink
at LN temperatures. It will thus be 9966 m when cold, leaving to
allow reasonably easy assembly of the shaft into the wheel by hand. To know
how immoveable the shaft will be, it is necessary to know the moduli of
elasticity of the components, as well as their geometry, and the interfacial
coefficient of friction.
Taking this example further, suppose that the wheel is much more rigid
than the brass and that the coefficient of friction, is 0.2, then, given a
brass modulus of E and ignoring Poisson ratio effects for simplicity, the shaft
will resist a 600 N or 60 kg axial force before it can be moved within the
wheel.
A detailed calculation for the case of a cylindrical shaft radius being
fitted with a ring of length L of outer radius gives the following result for
the required interference distance I (in metres) to give an axial grip force of
F, allowing for Poisson and for the wheel and shaft materials
Detailed calculations of interference fits are given in standard engineering
texts for simple geometry. There are effects due to the Poisson ratio
(generally, when a body is squeezed in in one direction, it squeezes out in
another direction) and effects due to anisotropy in elastic moduli. The force
with which an insert will grip also depends on how hollow the insert is and on
the inserted components thickness and elastic moduli. In some cases,
shrink-fitting will result in the materials elastic limit being exceeded,
reducing the tightness of fit and further complicating calculations as the
component will permanently distort slightly. In difficult shapes, numerical
modelling using a computer software package is possible, whilst for small
components there is no substitute for experiment. Finally, in production
lines, where it is required to fit many sets of components together without
individual measurement, allowance must be made for the worst cases of
smallest insert and largest housing (which may not grip well enough) and
largest insert with smallest housing (which may not be possible or may
overstress the components in the finished assembly).
Shrink-fitting is certainly convenient (LN-cooled components can easily
be handled with gloves, for example) and very inexpensive (a polystyrene
insulated pot or bath is the only item of capital equipment needed). Also,
LN rarely leads to any degradation of the surface finish of components.
Heating to a few hundred C is difficult to accomplish while keeping paint
finish intact and stronger heating may lead to oxidation staining as well as
giving a danger of unintentional heat-treatment effects. However, there are
some problems. Embrittlement of steels at low temperature may lead to
chipping or failure if a cold component is inadvertently dropped
61
or there
may be failure during warm up. Note also that some austenitic steels are in
effect heat-treated by LN exposure, which will usually not be desirable.
Cryotumbling and cryoblasting of rubber components. Here the embrittling
effect of rubber is employed to good effect. Small moulded components
normally have flash and sprues: superfluous material that is typically
removed by tumbling many components simultaneously in a large drum,
perhaps with the addition of ball-bearings, abrasives and water or other
liquids. After a few minutes or hours tumbling, the sharp edges, and all the
flash and sprues, have been removed and a smooth overall finish is obtained.
However, this process is difficult with rubbery materials unless liquid nitrogen
is added to the tumbling drum. With LN added, the physical properties of
rubber, hardness and brittleness, become similar to those of other hard plastic
and metal components, and tumbling of rubber becomes straightforward.
In a similar way, other processes can clean-up sprues and flash from plastic
components without the often undesirable rounding of edges that tumbling
entails. One such is the shot blasting of trays of LN-cooled components. It
should also be added that components from materials other than rubber can
usefully be deflashed with LN pretreatment. Small zinc die-castings and small
thermoplastic components can benefit from this method.
Concrete cooling. Concrete is made by binding together fine and coarse
aggregates (sand and stones, in other words) with Portland cement. Portland
cement is a mixture of silicon, magnesium and aluminium oxides that, when
water is added, reacts to produce an interlocking mesh of hydrated mineral
crystals forming a strong rock-like material. Unfortunately, the reaction is
exothermic and will not proceed correctly except near ambient temperatures.
In temperate climates, and where the thickness is not too great, ambient
cooling is sufficient. The correct choice of cement (exothermicity varies with
the cement oxide fractions) enables satisfactory concrete.
In hot countries the setting of concrete is seriously impaired, however,
because the achieved strength is lowered as temperatures are raised.
Cracking tends to occur in thicker sections because of the differential thermal
expansion of the mix during setting and the slow cooling rates of the centre of
thick-piece concrete. In addition, water tends to evaporate from the surface
of the concrete during setting before it has had time to react and set.
Structures needing high performance from concrete, therefore, are made
by cooling the concrete by adding liquid nitrogen to the mixing machine
during mixing. The cooled concrete is then laid or fed to the concrete pump for
placement. Cooled in this way, the concrete and its added water have
sufficient heat capacity to set correctly before the exothermic reactions
61
The author has found smaller steel parts still remarkably robust after cryocooling.
warm them up too much. Cracking and surface problems are avoided and
strength enhanced; by as much as 20% or more in some experiments.
Experiments have also shown that fears that the concrete will be seriously
weakened by freezing if there is a slight overdose of liquid nitrogen in the
mixer are unfounded. Freezing of concrete during setting generally causes
problems because the growth of ice crystals causes voids to be formed.
However, the freezing caused by LN injection is both fast and slight, so the
concrete setting is unaffected.
Cryogenic solvent/vapour recovery. Vapour recovery is where volatile
organic solvents, such as xylene or methyl cellusolve, used in cleaning or in
paints and coatings, are recovered for reuse. A vapour stream from dip
tanks, spray booths or drying kilns, following a coating or cleaning process,
is processed to recover the solvent and reduce environmental discharges.
The two drivers for the adoption of recovery systems are the increasing
environmental demands put on companies to limit solvent emissions and the
cost of solvent, particularly for large or continuously-operating plant.
A common system for recovery is to force air extracted from the vapour
source through carbon-based absorbers. The solvent may then be recov-
ered, or at least a sizeable fraction of it, by warming the absorber beds and
passing air (backwards) through them and then over a condenser. However,
this kind of simple bolt-on system needs stringent safety precautions, is
usually not very efficient and may be rather expensive to run. More
satisfactory installations can be achieved using a cryogenic coolant system
and inert gas.
A simple installation would take the extracted air stream from a plastic
tape drying kiln, and cool it, first using water-cooled surfaces, then using a
liquid nitrogen spray, typically onto a set of high surface area inert
metal-contacting surfaces. The result is a mixture of solvent and water, and a
much cleaner air stream to exhaust to atmosphere. The solvent and water
are separated, by distillation if necessary, and the solvent reused.
Instead of exhausting to the atmosphere, some coating processes can be
performed in a nitrogen atmosphere in a semi-closed cycle. An inert
nitrogen stream from the solvent exhaust cooler can often be used to provide
point-of-use inerting for a coating or cleaning process, reducing the effects
of atmospheric water and oxygen contamination (oxygen is undesirable for
some UV-set coatings, for example, and water causes problems with many
coatings) and eliminating fire and explosion hazards. Nitrogen is bled from
the system at a point, such as just before the point-of-use, where solvent
concentration is low and the nitrogen temperature relatively high. Such
systems are especially applicable to tape-coating processes in which the
process oven can be most effectively sealed.
The main cost savings attached to a solvent recovery system are related to
the ability to recycle the solvent. However, where a coating must be rapidly
Figure 3.36 Semi-closed cycle vapour recovery using liquid nitrogen injection.
evaporated by heating strongly (known as stoving) there is an additional
substantial saving to be made when using an inerted system. There is no
need to maintain a low concentration of vapour in the coating unit, so there
is no need to heat a large amount of air to keep the vapour concentration
low. Because there is no danger of ignition in a coater which is inerted,
vapour concentration can be much higher. In some cases (e.g. xylene)
solvent fraction can be raised from 0.5% or less (to stay below 30% of the
lower explosive limit) to as much as 30% by volume. If an oven maintained
at, say, 150C is converted from low vapour concentration in air to high
vapour in nitrogen, these savings can outweigh the cost of the liquid nitrogen
used and the solvent recovered is pure profit.
A semi-closed cycle system may allow higher temperatures to be used
safely, allowing for quicker drying of solvent from the workpiece and greater
production rates from a given size of production unit (Figure 3.36). Some
semi-closed systems achieve almost 100% solvent recovery as moisture
contamination is minimised, reducing undesirable solvent emissions dra-
matically. The potential for recovery is limited by the cooling achieved
versus the oven temperature. This is, in the absence of other effects, the
proportion represented by the vapour pressure at the cool temperature over
the vapour pressure at the oven temperature is the proportion recoverable,
i.e.
recovered (%)
On a typical exponential vapour pressure curve, VP = exp (E/kT) and the
recovery R is a function of the temperature difference and the
latent heat of evaporation (to which E is proportional)
With an oven temperature of 100C and an achieved cold temperature of
100C, for example, about 99.6% of octane will be recovered. With a
vapour content of only 0.4%, the vented gas resulting will be acceptable on
both environmental (VOC emissions) and safety grounds (at 0.4% by
volume, the mixture will not inflame in air). With achievement of
temperatures even approaching LN boiling point, 196C, even the 0.4% of
vapour escaping at 100C will be eliminated, giving an even more desirable
result.
A similar system for vapour recovery can also be installed as part of a tank
ullage inerting system in the large-scale storage of volatile materials such as
petrol (section 3.3.7). A system as shown in Figure 3.36 can be used on the
vent line from a tank ullage space. With petrol vapour at summer
temperatures constituting 5 or 10% of the gas in the nitrogen inert blanket,
the vent line recovery can be highly economic, with LN costs exceeded
several times by the value of the recovered volatiles. Vapour recovery
systems are now common in chemical tank farms both for recovery from
tank to tank pumping and also where tanks are off-loaded to or loaded from
haulage vehicles such as road or rail trucks, or transport containers.
With the intermittent nature of tank venting, it is rarely economic to
install a mechanical refrigeration plant of a large enough cooling capacity
and low enough temperature to carry out vapour recovery in tanks. A
system of low capital cost with a small LN tank but a high peak capacity at
low temperature is a much better match to tank venting vapour recovery.
The inexpensive carbon-bed-based systems are, however, competitive with
LN in many cases.
The nitrogen evolved in the recovery system is not directly usable in the
inerting of the tank for which the recovery system is operating since the
recovery process is needed when the tank is venting, not when it is intaking.
With a little ingenuity, however, where there are multiple tanks of similar
products, some of the evaporated LN can be used as (slightly impure)
nitrogen inert blanketing for other tanks. In this case, however, check
valves, flame arrestors and/or slam-shut valves triggered in the event of a
flame may be necessary to prevent a fire in a tank ullage space spreading via
the pipework system.
Cryomilling. As noted in section 3.4.1, spices and other foodstuffs tend to
heat up and clog the machinery in a high-speed bulk grinding operation. The
heat also drives off wanted volatile oils in spices and leads to other
degradation. The addition of LN to the grinding jaws removes unwanted
heat, adds an inert atmosphere and embrittles the product, leading to better
quality and higher grinding speeds.
The grinding of plastics and rubber is also now commonly done with liquid
nitrogen as these products are exceptionally difficult to shred or grind at
ambient temperatures, particularly where smaller particle sizes are needed.
Even where the particle sizes needed are large, as in shredding, there may be
big advantages in using LN grinding. Old tyres, for example, even those with
steel reinforcement, can be ground up. Their steel reinforcement causes big
problems for conventional shredding machinery but the cryogenic process
allows the steel reinforcement to be easily separated from the rubber itself.
In grinding tasks such as tyre shredding, there is an additional effect that
comes into play with the use of cryogenic embrittlement: differential
embrittlement. Although two materials may both be embrittled at LN
temperatures, one material will be affected more than the other. On
crushing or hammering, one material is broken easily into small crumbs,
whilst the other only breaks to a limited extent. This can be a big help with
separation and hence with recycling.
62
The effects of differential embritt-
lement are used in commercial practice, for example in removing plastic
insulation from cables whose copper or aluminium scrap content is to be
reused.
Although energy is needed to create the LN, the energy needed in the
shredding machinery may be high compared to that required for crumbing in
a cryogenic process. Overall, therefore, cryogrinding and crumbing can be
more energy efficient. The availability of LN grinding has even led to new
products. In the surfacing on childrens playgrounds, for example, small
chips of rubber are now common. They provide a desirable degree of
resilience beneath play apparatus where a fall on to concrete for a child
could be serious. Made by crumbing with LN assist in many cases, the chips
can be left loose or bonded with a small amount of adhesive to provide a
resilient equivalent of tar macadam road surfacing. A similar process adds
crumbed rubber to stone chippings in road surfacing, changing the roads
grip characteristics.
Plastics are now commonly crumbed cryogenically to provide a con-
venient, free-flowing feedstock for moulding and other plastic processing
operations. Plastics crumbed with ambient shredders often have crumbs
with a tail which cause the aggregate plastic to clog and clump. Plastic with
pigments, as used in, for example, xerographic toner, can conveniently be
prepared by cryogrinding of bulk material, with powdered plastic of almost
arbitrarily small particle size down to a few microns being available. This is
impossible by other methods. Although cryogrinding on a small scale often
employs the simplest equipment, simply adding a fine spray of LN to a more
or less normal grinder, other systems employ a counter-current screw feeder
to maximise the efficient use of the cold potential in the LN and achieve a
good level of inerting.
62
The plastic goes below its glass transition temperature and embrittles, whilst the copper or
aluminium remains ductile. Zuwala et al. (1994) describe the ribbed rollers needed to cause
crumbing of the plastic while preserving metal wires.
Throughput of a product in grinder is often greatly increased by the
addition of LN cooling and, as noted in section 3.4.1 above, problems with
overheating (partial melting of plastic chips, for example), oxidation
damage and explosion/fire hazards can be avoided. Grinders of types which
are normally useless on soft material, but are desirable in that they have low
energy consumption (such as centrifugal hammer-mills or ball-mills, for
example), are often usable with materials embrittled by LN.
Freeze-drying. Freeze-drying or lyophilisation is a technique for the
preservation of organic compounds susceptible to degradation if simply left
in the ambient atmosphere. Lyophilisation is mainly applied to pharmaceu-
ticals and other high-value products, including some high-value foods,
although food materials such as coffee and soup are also freeze-dried in
some quantities but using a simpler and less expensive process. Whereas
normally aqueous or hydrated products are dried by heating gently and
blowing air copiously through or across the medium, freeze-drying removes
water by first freezing and then removing the ice formed by sublimation.
Roughly speaking, whereas normal evaporation results in a product
shrinking by an amount equal to the water loss and maintaining its density,
freeze-dried products have about the same volume as the original product
since the ice crystals keep the structure of the product intact, leaving
interstices as the ice sublimes. The interstices allow the rapid rehydration of
the product when it is ready to use and this rehydration largely reconstitutes
the original structure. Evaporated products often rehydrate poorly since the
water cannot penetrate the shrunken, shrivelled layers of solid product and
the heating during evaporation often alters the chemical constitution of the
product. Freeze-dried products made by freeze-drying of soluble com-
pounds form granules composed of loosely bound microcrystals that are
more readily dissolved than normally evaporated crystals, which are much
larger and cemented together.
Rapid initial freezing of the product can be helpful as the freezing process
itself can cause destruction of some products, for example plant cell walls,
because of the growth of large pointed ice crystals. On the other hand, a
slower freezing with larger crystals leads to a speedier drying process since
there will be a more open structure in the product with better vapour paths.
Most lyophilisation systems use mechanical refrigeration at approximately
50C, often by simply placing the products on cooled trays in the vacuum
chamber before applying vacuum. However, with thin products where
vapour transit is not an issue, or where an extended drying cycle can be
allowed, liquid nitrogen freezing is advantageous. The vacuum pressure
applied is 1 mbar or less, typically for a period of several hours or more
whilst no further cooling is applied but heat is allowed to enter the vacuum
chamber so that the temperature does not fall too much.
After initial removal of water, more water can be driven off by gently
warming the product back towards room temperature over several hours or
more whilst maintaining the vacuum. Once the product has warmed up back
to room temperature it will be found to comprise a porous, rather friable
solid, typically hygroscopic but containing just a small percentage of water.
At this level of water most materials will, if kept in sealed containers under
dry nitrogen or air, be stable against chemical degradation and, more
especially, be kept free from fungal, bacterial or other spoiling action. Large
batch lyophilisers will accept up to a tonne of wet hydrated product for
processing, in thicknesses up to approximately 1 cm.
Plastic-moulding cooling. Here liquid nitrogen is employed to cool a
freshly moulded thermoplastic item so that it can be removed from the
injection moulding machine more quickly. In this way, the productivity of
these machines can be enhanced, sometimes by as much as 100%.
Both blow moulding and injection moulding can be speeded up. In a
typical blow-moulding application, a gob of plastic is inflated by air pressure
until it fills up the mould, which might be as as small as a 10 g plastic bottle or
as big as a 100 1 fuel tank. As soon as the parison, the inflated gob, has
contacted the walls of the mould, its outer surface begins to cool on the
water-cooled mould walls. At this point a brief but precise stream of LN is
sprayed inside, cooling the inside of the parison. With cooling mostly
complete, the parison is solid enough for it to be ejected from the opened
mould much more quickly than would otherwise be the case.
Early plastic-moulding cooling with LN was somewhat hit and miss.
Rejects occurred frequently, particularly during start-up of production,
because LN in the cryogenic feed line filled with vapour and insufficient LN
was admitted when the cryogen valve opened. Ways around this include
careful designs of valves, use of a cool-down sequence, so that the
components reach a steady-state temperature, and the installation of a
phase-separator immediately above the moulding machine. AIRCO in the
USA is trying the use of a dipping pipe in place of a valve. The dipping pipe
lies normally in the vapour space of the phase separator tank. When LN is
demanded, the short dipping pipe is immersed in the LN and pressure feed
forces a known volume of gas followed by cryogen into the mould. At the
end of cooling, the dipping pipe is raised to drain it of LN before shutting off
the cryogen (Figure 3.37).
Extrusion cooling. Just as moulding can be speeded up by cryogenic
cooling, extrusion processes for plastics (and some metals such as alu-
minium) can be speeded up, and often quality improved, by the use of LN
cooling. Particularly with higher temperature extrusions, the inertness of
the nitrogen atmosphere can improve surface quality. The extrusion process
is speeded up because a higher temperature can be used in the feed material,
Figure 3.37 Dipping-pipe technique for accurate dosing of liquid nitrogen (courtesy of
AIRCO).
giving a lower viscosity and hence higher throughput for a given die and
extrusion pressure. In other installations, production is not limited by the
extrusion screw but by the rate at which the product can be cooled
sufficiently to be stored or further processed. Again, LN cooling can reduce
this bottleneck.
LN cooling can be added to existing equipment for minimal capital outlay.
This sort of installation is often encountered as an extra to inexpensively
speed up the output of a previously installed piece of capital plant. However,
if careful calculations are done, LN often proves to be economic from the
outset.
Blown-film extrusion is to some extent a combination of blow moulding
and extrusion is commonly used to make thin polyethylene film (Figure
3.38). In this process, a plastic tube is extruded and, while still hot, inflated
with controlled internal air pressure to twice or more its starting diameter.
The inflation pressure is very small, centimetres of water gauge or less, and
must be carefully controlled. It should be varied with the extrusion speed
and take-up roll speed to provide the correct amount of inflation and thus a
film of the desired thickness. The film cools in the air sufficiently that it may
then be led to a V-shaped funnel with a pair of take-up rollers pulling the film
through the slot in the V-funnel. The result is a roll of film in the form of a
wide, thin-walled tube, neatly folded at each edge. When slit up one side,
the film roll is thus in length only half the width of the opened out film,
making it a convenient form for handling and selling. The film extrusion rate
can be increased by circulating cool air on the inside of the film (in the
normal system, the only cooling is on the outside of the film). Systems using
LN to make cold nitrogen to cool a partial closed-circuit air-cooling system
on the inside of the film have been shown to boost production rates by
approximately 50%.
Cryogenic fog and theatrical effects. There are serious uses for large flows
of liquid nitrogen to simulate cold upper atmosphere conditions. The
combination of liquid nitrogen and water vapour can be controlled with
suitable apparatus to yield thick, cold fogs, which are also useful in testing
instruments and equipment for use in foggy conditions.
Liquid nitrogen can also be used for making steam jets or fountains as a
stage effect. The pressure of the liquid itself, boiling, can be used to propel it
through small jets. The liquid stream issuing from the jets quickly boils into
cold gas which condenses moisture in the ambient air, creating a spectacular
but reasonably safe fountain. Similarly, by allowing the LN to boil before a
larger orifice, a jet of steam effect can be created.
foggers are a more standard stage effect. Water and dry ice are
mixed, resulting in the copious evolution of cold and water droplets.
The mixture, even when diluted with much air, is considerably heavier than
Figure 3.38 Blown film plastic extrusion equipment. Molten plastic extruding from a ring
(centre) is inflated and rapidly cooled by cold air to form a continuous tube. The tubing is
thoroughly cooled and then wound onto reels in a continuous process. Cryogenic nitrogen
cooling in such equipment allows increased productivity and the use of polymers normally not
suitable for the process (courtesy of Battenfield Gloenco Extrusion Systems Ltd).
ambient air and forms a fog layer which flows over the floor. A more
long-lasting fog effect can be achieved by using dry ice or LN to cool the
air/smoke mixture from a glycol-based stage smoke machine so that the
smoke/air stays in a layer near the ground.
3.7.4 Dry ice bead-blasting
An application which makes ingenious use of the properties of solid
pellets is dry ice bead-blasting. Ordinary sand-blasting, using sand driven by
compressed air, is a very useful technique for preparing surfaces for painting
or removing old paint from metals. Bead-blasting uses tiny manufactured
beads of glass or quartz, allowing a greater control over the process and
reducing the contamination of the workpiece by tiny sand particles. The
beads can be recovered and reused to some extent.
Unlike sand-blasting, with all its problems of apparatus wear and heavy
contamination of the working area with sand, pellets are very clean.
After they have impinged on the surface being treated, they simply sublime
into the air. If humidity is kept low, then condensation of atmospheric
moisture can be minimised. The technique is only limited by the necessarily
high cost of the process; the beads, which are not recovered, and the
additional equipment are expensive. Paint-stripping of aircraft is one
application which justifies the extra costs.
3.7.5 Liquid helium and NMR/MRI
The outstanding exception to the rule that liquid helium cryogenics is of no
interest to industry is, of course, the use of LHe-cooled magnets for nuclear
magnetic resonance (NMR). Although ordinary copper electromagnets can
be used for any chosen field strength and size, at high field strengths and
volumes (say above 0.1 it becomes economical to employ a
superconductor cooled in liquid helium. There are two principal appli-
cations for nuclear magnetic resonance: research on molecular structure
(NMR) and human body imaging (magnetic resonance imaging, MRI).
NMR scanners are generally required to have a field in excess of
approximately 1 tesla and to have a volume sufficient to contain the patient
being scanned. Although it is possible to use lower fields, signal-to-noise
ratios rise as a power of approximately 3/2 so the higher fields possible
without power penalty using superconductors are desirable. The highest
fields are used for molecular structure research.
NMR and pharmaceutical research. NMR can be used to understand the
chemistry of complex molecules. Many atomic nuclei have a magnetic dipole
value. The normal isotope of hydrogen and the rarer stable isotope of
carbon
I3
C are the most common examples. When liquids containing these
atoms are placed in a magnetic field and inside an RF oscillating electric
field, resonance effects are seen.
The nuclear magnetic resonance frequency of a nucleus depends on the
magnetic field at that nucleus. In effect, the nucleus can be considered to be
like a little magnetic spinning top, whose precession frequency will depend
on the applied magnetic field tending to twist its axis of rotation. The applied
field, however, depends not only on the applied field but also on the internal
magnetic fields caused by the magnetic dipole moments of the circulating
electrons in the molecule in which that nucleus is sat. Different molecular
Figure 3.39 NMR scan showing several different hydrogen nucleus resonances from different
parts of a molecule of ethanol at low (a) and high (b) resolution.
environments lead to slightly different frequencies of resonance and it is
these differences which are measured in order to infer chemical structure.
These differences in resonance frequency grow linearly with magnetic
field and can be clearly seen only at fields of several tesla. As the desire has
arisen to see more subtle molecular effects, fields have had to increase until
today the leading-edge equipment is using heroic fields of up to approxi-
mately 20 tesla. Higher fields still are needed in order to achieve the
maximum resolution in the observation of the environment of each of the
NMR active atoms.
Above 20 tesla there is intensive work going on to use ceramic
superconductors as an inner booster solenoid. These booster coils will most
likely also be cooled to 4.2 K by liquid He. The material is used not
because it offers a high superconducting transition temperature but
because superconductivity in many materials persists to very high
magnetic fields. As such booster coils become available, researchers will be
able to carry out NMR biochemical strucutural analysis at higher resol-
ution.
NMR for MRI body scanner magnets. MRI units are designed to image
the inside of the human body non-invasively by using nuclear magnetic
resonance effect. The part of the person to be scanned must be placed
entirely within the scanner so large magnets are required.
As for NMR, the nuclear magnetic resonance frequency of a nucleus
such as hydrogen depends on the applied magnetic field. However, by
arranging a field which varies slightly across the person being scanned,
hydrogen atoms in different spatial locations will differ slightly in
resonance frequency and those different parts may be addressed by
tuning the detector/transmitter coil, the intensity of resonation and the
decay time reflecting the concentration and chemical state of the
hydrogen atoms there.
By means of a computer, these many different parts (planes) are
addressed sequentially, building up a tomograph, i.e. a set of data
representing the density of atoms in certain planes or lines. This
tomographic data can then be processed by computer into a three-
dimensional data set representing the density of atoms versus x, y and z
coordinates in the scanned persons body. This can then be displayed in
clinically useful form as, for example, cross-sections in different planes.
The NMR body scanner business has grown to be the biggest
application for liquid helium and superconductors, with every large
hospital in the developed world running them. This would be surprising
to a time traveller from, say, 1965, when the predicted main applications
for superconductors were power transmission, superconducting levitating
trains or superconducting motors. (Examples of all these applications
exist but only in prototype form; they have not achieved any commercial
success.)
3.7.6 Gaps in the cryogenic spectrum
There are gaps in the cryogenic spectrum, as noted in section 3.7.1
between LHe at 4.2 K and LN at 77 K, and between liquid argon at 87 K
and carbon dioxide at 195 K.
In the lower gap there is only neon and hydrogen to choose from. Neon
is one of the most exotic cryogens ever used and has been used only in
satellites. It is used in satellites not as a liquid but in solid form. Neon is
unusual in its small liquid range (2527 K) and the large amount of gas
evolved on vaporisation (1410 volumes gas/volume liquid). Its convenient
boiling/melting points and reasonably high latent heat capacity make it a
reasonable choice for cooling satellite detectors in the region of 30 K.
Without a suitable boiling liquid, cooling can be carried out by careful
control of a lower boiling cryogen, often liquid helium. However, this is
very expensive; not only is LHe many times more expensive than LN but
its heat capacity per unit gas volume is 64 times lower. In recent years,
this practice has been largely supplanted by the use of direct gas
refrigerator cooling systems, typically Stirling cycle expansion engines
using helium as the working fluid. To provide a steady temperature for
even a small heat load requires a surprisingly large cooling engine,
however, because of the low heat capacity of gases.
There are in fact few problems today in operating small heat load
systems in these gaps using gaseous cooling. However, gaseous cooling is
not used unless necessary. A boiling liquid bath provides a large heat
capacity and a steady temperature which are difficult and expensive to
provide with gaseous systems.
In the higher gap, there are many compounds with reasonable boiling
points, although many of them are unsuitable on other grounds, e.g. cost,
(krypton, 120 K), flammability or reactivity (methane, 112 K or fluorine
monoxide, 128 K). The refrigeration industry swiftly adopted the CFCs
once industrial fluorine chemistry had enabled their low-cost production,
because they enabled safe operation of cooling plant near and beyond the
top of the gap range. The lowest boiling CFCs are useful cryogens; they
are mostly very inert, nonflammable and have useful heat capacities.
However, the concerns for the ozone layer noted in section 2.5.12 will
prevent their use. Compounds such as the HCFCs and others with a low
ozone depletion potential (ODP), such as carbon tetrafluoride (halo-
carbon 14, boiling point 145 K), however, will continue to be widely
used.
3.7.7 Cryogenic pipelines
Although now entirely practicable for long-distance liquid natural gas
(LNG), cryogenic pipelines are nevertheless almost unknown. A study by
Berbesson (1979) snowed that the economic factors for transporting
natural gas as a liquid are quite favourable. A 1050 mm diameter 100 bar
steel gas pipeline could be replaced by a 500 mm diameter low-pressure
nickel steel liquid pipeline, resulting in savings of material and pipelaying
costs. Other savings occur in running the pipeline because the cost of
pumping the gas in liquid form is less than half that of pumping gas itself.
Fortuitously, the strength of the nickel steel increases at a low
operational temperature, enabling a thinner material to be used or
enabling a longitudinally fixed line to be used. The latter feature avoids
the need for compensation of the thermal contraction of the line as it
cools down. Interestingly, the insulation needed for the line is a simple 75
or 100 mm layer of foam; the principal heat load in a high flow rate
pipeline would be due to friction losses. The relative ease of construction
of a methane line contrasts strongly with the complexities of the
LHe-cooled pipeline needed for cryogenic electric power trans-
mission.
3.8 Gases across the breadth of industry
3.8.1 Gases in space
Rocket propellants. Large rockets are propelled mostly by combusting a
liquid fuel with a liquid oxidiser. The fundamental parameter of importance
in a rocket fuel is its specific impulse, I, which has the dimensions of time.
The specific impulse is the thrust (in mass units at 1 g) developed by an ideal
adiabatic rocket divided by the mass flow rate of the fuel/oxidiser consumed.
The reason why this is the standard measure of a rocket fuels efficiency is
that it gives the maximum change in momentum that can be added to the
rocket for the expenditure of a given mass of fuel. If a mass of fuel is
burned in time t, then thrust F/g will be produced for time t, giving for the
rocket of mass as follows
where V is the rocket speed and g is the acceleration due to gravity. The
specific impulse is dependent on the adiabatic temperature achieved in the
fuel/oxidant reaction and the molecular mass of the ejected products of
reactions. The exhaust velocity v is given by so
where m is the mean molecular weight of the products of reaction.
A number of oxidants have been tried and many fuels. The most efficient
are rather expensive, considering that each rocket fuelling requires many
tonnes of fuel, so cheaper combinations are more popular. The use of
LO/diborane, for example, would be at least 100 times more expensive than
LO/paraffin. Examples of the specific impulse for some gas mixtures are
given in Table 3.13.
The only high-performance propellants which are in use today in large
quantities are LO/paraffin and LO/LH. They have the advantage of:
comparatively good performance
low cost
innocuous by-products
controllability.
Although liquid hydrogen does present some handling problems, it and
liquid oxygen are now routinely used on every space launch. In recent years,
liquid hydrogen has rarely led to problems. Accidents such as the early
Ariane rocket losses were due to motor problems, whilst it was solid fuel
motors that caused the NASA Space Shuttle Challenger disaster.
It is interesting to note that, on energy per unit weight grounds, hydrogen
is the best source of power for an air-breathing vehicle of any kind. It is
about three times better than petroleum fuels in an engine:
A Boeing 747 Jumbo Jet airliner, for example, fuelled on liquid hydrogen
but incorporating no other modifications, could in principle fly around the
world non-stop without refuelling.
63
The example does, however, ignore the problem that would occur with
the bulk of hydrogen, which, with LH density around is very
high. Liquid/solid hydrogen slush achieves a higher energy density than
liquid hydrogen and has been used on occasion to slightly uprate the
performance of space vehicles.
Spaceborne supplies. Gas technology provides numerous critical elements
in the world space programme. Clearly the massive usage of liquid oxygen,
and to a lesser extent liquid hydrogen, dominates the gases used in space.
However, significant smaller uses for gases have also been found to be
essential. Hydrogen and oxygen are used for generating electric power from
fuel cells, helium for scientific experiments needing superconductors and
hypergolic propellants are used for manoeuvering. Compressed gases, such
as nitrogen or are used for small orientation manoeuvres and for
breathing atmospheres needed for human pilots. The unfortunate and
near-fatal explosion of the tank on one of the Apollo moonshots served
63
This would give an airlines round-the-world ticket an entirely new meaning; a latter day
Phineas Fogg would finally not have to worry about missing his connections! A round-the-world
non-refuelled flight was achieved as a record-breaking attempt with conventional fuel, flown by
an American couple (the Rutans) in a strange glider-style aircraft.
to emphasise how dependent spacecraft are on these many different gas
systems.
3.8.2 Cleaning: thermal blasting of stone
Smooth granite surfaces, carefully polished, are hardwearing and not
particularly expensive. Granite and similar stones are now popular for
pavements in indoor and outdoor pedestrian areas such as shopping malls.
However, when wet, the granite, particularly in heavily used areas where it
is particularly smoothly polished, is slippery.
One answer to this it to treat the granite surface to make it slightly rough.
Although grinding can achieve this, a neater solution is to use the intense
heat of an oxy/acetylene flame to momentarily heat the surface of the
granite. The surface layer expands, whilst the cooler body of rock remains
unaffected. The resultant stresses cause the surface to spall off. As little as a
millimetre can be removed with care. The surface left behind is fairly evenly
rough and non-slip but is still attractive to look at.
3.8.3 Firefighting
Several gases have come to prominence in fire extinguishing. Most gaseous
firefighting works simply by excluding oxygen from the air by displacement
with an inert gas. The brominated fluorocarbon contained in bromochloro-
difluoromethane (BCF) extinguishers is a more ingenious choice of gas
because the bromine released when the gas contacts a flame actually
interferes with the mechanism of combustion; it acts as a negative catalyst. A
small quantity of BCF is equivalent to a much greater volume of or N
2
.
However, CFCs and brominated CFCs are serious greenhouse and strato-
spheric ozone-destroying gases, and they are the subject of severe restric-
tions on production. Hence, only two gases are likely to be important in
firefighting in the future: carbon dioxide and nitrogen. A recent patent from
the fire prevention company Wormald International illustrates an appli-
cation for both gases together. They claim that a 40% Ar, 8% 52%
mixture is ideally suited for automatic extinguishing of fires inside indoor
equipment such as computers. The mixture is inert and suppresses the fire
but any gas which leaks out has the same density as air and will not be
displaced upwards or downwards. Also, the content will stimulate the
breathing of people near the fire, enhancing their ability to survive even with
lower concentrations of oxygen in the air locally.
Both and are used as propellants for dry powder extinguishers
typically discharging sodium bicarbonate. On its own, is a convenient
and effective extinguishing agent for small hand-held or wheeled-trolley
extinguishers. These can be used on any fire with no danger of electric shock
if electrical apparatus is in the fire or near to it, and they will not damage
equipment. However, fires extinguished, particularly hot solids on fire, can
reignite when the stream is switched off as the gas will not cool a
conflagration as much as water. The addition of dry inert powder to the gas is
useful because the powder will coat the burning solid, helping to prevent
reignition.
The most commonly used gas in large fires is nitrogen because of its ease of
vaporisation and handling. Ships holds and coal mines are examples of
peculiar circumstances where the use of nitrogen can be a potent weapon in
firefighting. For an inert gas to be useful on a big fire, the seat of the fire must
be more or less sealed against ingress of more air. Huge quantities of
nitrogen can be needed to tackle fires effectively, however, as although they
are more or less sealed, many ships holds and underground workings have
huge volumes. Many mines have tunnels several miles long, for example.
The great advantage of nitrogen is that it will not damage equipment or
lead to flooding. Coal mine fires are very common; methane, carbon
monoxide and finely powdered coal are all flammable hazards. The placing
of heaps of chalk on ledges throughout coal mines is still carried out to
suppress air/dust explosions with coal. When an explosion starts, dust is
shaken from the ledges. Instead of more coal dust being shaken down to
exacerbate the flame, chalk dust appears and forms a fire-suppressing cloud.
Flooding is always an option with mine fires but flooding of the mine shafts
and tunnels can easily turn out to be irreversible as it damages equipment
and can block the workings with mud.
Nitrogen-filled soap/water foam has occasionally been used to tackle fires
in unusual industrial or marine situations. More commonly, foam inerting is
used during demolition of tanks, pipework and chemical plant used for
flammable liquids (the procedure is described in section 3.3.7).
3.8.4 Gases in tyres and inflatables
Inflatable emergency lifejackets and inflatable boats are mostly inflated
automatically from small cylinders of liquid Carbon dioxide cylinders
are much lighter than a cylinder with a similar volume of nitrogen. There are
potential problems, however, with diffusion gas loss and nozzle blocking.
is lost by diffusion through plastic and rubber faster than air or nitrogen
but this sort of slow loss of inflation is not important in emergency
applications. can freeze solid in the nozzle of the inflation cylinder.
However, careful sizing of inflation system components avoids this.
Nitrogen can be advantageously used for filling tyres. Tyres under
arduous conditions of use (at high speed and under heavy load) tend to heat
up. The rubber on the inside of the tyre tends to decompose to some extent
in these circumstances, being oxidised by the hot compressed air within it.
Degradation of a nitrogen-filled tyre is slowed by the absence of oxygen
inside it, giving a longer lifetime. Also, momentary ignition inside tyres,
caused by the formation of a flammable mixture of high-pressure air and
rubber particles is not unknown. This is prevented by Aircraft tyres are
almost invariably filled with and the practice is spreading to truck tyres.
Automobile tyres have very light loadings, low pressure and low heat build
so little is probably gained by the use of
64
3.8.5 Gases in glass manufacture
Oxygen is sometimes used in glass tanks for flame enhancement. The
productivity of a glass tank can be enhanced if its charge can be heated to
melting point more quickly. Most of the cost of glass production is in the
melting of the materials at the start of the process and small improvements in
this stage yield big advantages in reduced energy consumption. The
materials used do not have high thermal conductivity and have high melting
points. Furthermore, carbonates of calcium and magnesium used in glasses
must be calcined during the melting, driving off carbon dioxide during the
melt. High temperatures and a very high heat flux are needed. These are
only obtainable from specialised electric furnaces or specialised gas-fired
furnaces. Electrical melting is used in small tanks but is uneconomic for large
quantities of glass. All this means that melting is ideally carried out with an
extremely hot, highly radiating flame.
Unfortunately, the emissivity of a standard gas flame at the most effective
absorbtion bands of the glass is not high. Traditionally this has been
approached, along with the need to conserve energy from exhaust gases
from the melting furnace, by the use of recuperative or regenerative
furnaces (section 3.1.9). Most typical is the use of checkerwork or hollow
grid brick assemblies that are warmed by exhaust gases and then switched to
heat incoming combustion air. The combustion air can be heated to between
1000 and 1300C, allowing the glass charge to be heated to approximately
1600C with ordinary natural gas and air. However, huge assemblies of
bricks are needed and the switching of hot gas streams is not easy, giving rise
to undesirable fluctuations in furnace temperature.
An alternative is -enrichment of the furnace air, giving a hotter flame
that has a higher infrared emission for the glass to absorb. enrichment is
controllable, gives steady temperatures, can be used to enhance existing
furnace output in retrofit installations and is lower in capital cost. Many
64
The use of sulphur hexafluoride has been suggested. The potential advantages of are not,
however, entirely clear. It is an inert gas, so would prevent oxidation and it might enhance
convective cooling within the tyre because of its high molecular weight. It is expensive,
however.
advanced glass processors are now using oxygen-enriched flames. -based
plants can be smaller and lower in capital costs than regenerative plants,
whilst existing plants can sometimes be stretched in capacity by the use of
oxygen-injected burners or by oxygen enrichment of air supply. The amount
of fuel used can be lowered since the hotter oxygen flame radiates very much
more brightly (following the radiation law) and transfers heat to the glass
more effectively.
Keeping the flame temperature as before but enhancing the emissivity of
the flame by seeding it with other gases is another approach. Acetylene has
been tried in this role with some promising results. Flat-flame oxy/fuel
burners (burners arranged to produce a horizontal plate-shaped flame) are
being tried in glass-tank heating and there are claims that the flat shape
further benefits heat transfer to the glass charge.
Glass furnaces have problems with emissions. The high temperature
of the flames needed means that they have much higher production
than most industrial furnaces. The use of pure oxygen, which avoids any
production simply by excluding rather than air may be the solution
ultimately. However, many current furnaces have to solve their
problems by processing downstream. One of the common techniques is that
of selective catalytic reduction (SCR). In SCR a carefully metered stream of
gaseous ammonia is added to the exhaust stack at a point where it is still very
hot, approximately 350C. The result is a reaction yielding harmless
nitrogen
Without the catalyst, higher temperatures are needed and produc-
tion occurs. The amount of liquid anhydrous ammonia needed is rather
modest since levels are in the tens of ppm range and the injection
equipment needed (an ammonia vaporiser and a metering valve, perhaps
adjusted by a servo operated by an analyser) is small. Various transition
metal oxide catalysts are in use and there is currently active development
work on improving these, particularly on reducing the temperature at which
the SCR reaction can be undertaken. British Technology Group (BTG) in
London together with a Dutch group, have one such improved catalyst.
Their patent (Kapteijn and Singoredjo, 1993) describes a process of applying
manganese salt in a solution to a porous alumina support medium, at a small
percentage loading of manganese, then calcining the catalyst to 400C for a
few hours. They claim useful activity for this catalyst at 100C with
commercial operation proposed to run only a little hotter than this.
Float glass. The production of glass in flat, optically perfect sheets has
been pursued for millenia. Only in the last 30 years, however, has a low-cost
process been developed and widely adopted for quality glass: the float glass
Figure 3.40 Float glass process.
process (Figure 3.40). The modern float glass process involves the addition
of heat, raw materials and scrap glass (cullet) to a large tank containing up to
hundreds of tonnes of molten glass at up to 1600C. The glass is drawn off
continously over a weir at the cooler end of the glass tank and run out at
1100C onto the surface of a bath of molten tin, 4 m or more wide. As the 4 m
wide strip of glass cools to 600C while moving across the tin bath, it solidifies
and can be drawn off by rollers downstream. There then follows a long
(100 m) row of rollers. The continuous sheet of glass cools slowly in this
section (the annealing lehr) to around 100 or 200C. Cutting and handling
can then be carried out. The production output of a float glass plant is
prodigious, 500 tonnes per day or more, and computerised automated
cutting stations and warehouses are typically needed to handle it.
The formation of the slightest quantity of dross on the surface of the
molten tin bath would give rise to imperfections in the glass produced. The
dross can be prevented, as in soldering, by the use of a mixture.
Glass bottles are mass-produced by a blow-moulding process not
dissimilar to that used for plastic bottle manufacture (section 3.7.3), except
that the glass is much hotter. However, the benefits of faster production as a
result of internal cooling could also be obtained in blown-glass products by
using a cryogenic nitrogen blast on the inside of the parison.
3.8.6 Coated glass
The first coated glass was produced by chemical processes such as the silver
mirror reaction used in chemistry as a test for aldehyde compounds.
Transparent or semitransparent coatings are difficult, though not imposs-
ible, to make in this way. Tin oxide-coated glass is transparent but
electrically conductive and can be made in a simple way by heating glass in
an atmosphere of vaporised Many glass coatings are in fact carefully
glued polymer films. Some safety glass is made in this way. Thin polymer
layers can peel off over a long period, however, are not completely UV
resistant, and are soft and easily scratched. True coatings of hard metals
and minerals are more satisfactory, in general, but more difficult to
achieve.
One of the most useful coatings for glass is IR active low-electron
multilayer coating. Glass with coating, when used in conjunction
with double glazing, allows sunlight in but does not allow heat energy out
of a building. It is an important energy conservation measure in cool
climates. The coating has alternating layers of high and low refractive
index spaced at intervals of a quarter wavelength thick so that light of
typical room temperature wavelength (5 to is reflected from the
different layers and the reflected waves interfere constructively. The
overall result is that the glass has a low effective emissivity in the infrared
region of the light spectrum; even when it is hot, it will not radiate much
light energy. If it is installed on the inner side of the inner pane of double
glazing, it will prevent heat transfer by radiation from the inner pane to the
outer pane. (The layer will also work if installed on the inside of the outer
pane as it will there act to reduce the absorption of infrared emitted from
the inner pane.) Visible light is transmitted almost unaffected because light
reflected from the different layers does not interfere constructively. Similar
multilayer coatings are also used in tropical climates where the low
emissivity also means that heat energy is reflected away. Tropical
glass incorporates an overall grey tint to reduce heating from visible solar
irradiation.
Although coating can be achieved with a polymer film, a more
satisfactory hard coating is achieved by using silane/oxygen-based gas
mixtures to deposit silicon dioxide-based CVD layers on glass in a similar
way to the CVD glassy layers used in semiconductors (section 3.5.6).
These layers can achieve simultaneously high visible transparency and
optical quality, low IR emissivity and a reasonably durable, washable
surface. The Pilkington Kappafloat glass is of this type and
glasses such as Kappafloat are now widely used in energy-saving buildings.
A recent patent reveals some of the difficulties of the coatings. The
composition of ordinary float glass is dominated by sodium and potassium
silicates. As in semiconductors, the alkali metal ions are relatively mobile
and will migrate from the glass substrate and bleed through into the
circuitry layers below. In the case of fine glass used for liquid crystal display
(LCD) screen manufacture this diffusion of alkali ions can kill the circuit.
This problem can be prevented by using an oxygen-containing gas which
does not react with silane (or carbon dioxide) and an unsaturated
hydrocarbon mixed with silane to deposit the first layer at 600C. The
resultant 50 nm layer of quartz-like glass is sufficient to suppress alkali
migration. IR-coated glass and other types have for some time been made
on glass straight from the float-glass tank. Until recently, fine-coated glass,
such as is required for LCDs, has been produced only off-line on cold glass.
More patent applications reveal that the aim of some glass manufacturers is
to produce LCD quality fine coatings on hot glass in 4 m widths.
Sputter-coating of glass. Thin-film processes using low-pressure argon
gas-based sputtering are often nowadays used for coating glass with metals
(section 3.5.6). Vacuum-deposited aluminium is used for high-quality
mirrors such as the front-silvered mirrors used in lasers. Extremely thin
gold-sputtered film made in this way has a brown appearance and provides a
useful barrier to unwanted sunlight, and sun heat, in air-conditioned
buildings with large glazed areas. Solar gain, the heating of a building due
to the suns heat through its windows, is often nowadays controlled by
coated glazing. Car windscreens are also commonly coated using a
sputtering process. The substrate is either heat-treated (toughened) glass or,
for windscreens, more common is a laminate of the Triplex type, i.e.
glass/polybutyrate/glass. Glass can be sputter-coated with simple opaque
metal/metal oxide layers which act as absorbers or reflectors on the glass
providing, for example, sun-shade or reduced overall transmission.
An argon (sometimes helium) low pressure (approximately mbar)
glow discharge sputter process is typically used. Although the principles are
the same as for semiconductor sputtering machines, glass sheet sputtering is
performed in enormous machines at very high speed. The sputter target in
machines for flat glass is often not a fixed target but a rotating cylinder.
Magnetic confinement of the sputtering ions (magnetron sputtering) is
employed. The magnets are mounted inside the rotating cylindrical target,
enabling higher powers and higher coating rates to be used and allowing
good utilisation (over 50% or more) of the sputter target material. A large
system will use around 800 V on the target, attracting the positive argon ions
to sputter off the target. Very high speed coating is possible (several
hundred square metres per hour or hundreds of hectares per annum) (see
Figure 3.4.1).
A more sophisticated automotive glass product is the conductive winds-
creen. Simple metal grid patterns are used for rear windscreen demisting
applications but for front windscreens the metal lines are unacceptable. For
demisting service, front windscreens can be coated with a simple conductive
oxide layer such as tin oxide or indium tin oxide (ITO), applied by chemical
deposition of vapours such as followed by oxidation in air. However,
tin oxide and ITO layers, although both conductive and transparent, have a
high resistance and only a small amount of power can be conducted through
them; sufficient perhaps for demisting but inadequate for deicing. The
approach taken by the AIRCO Solar glass-coating firm, starting 10 years or
so ago, has been to use metallic layers of high conductivity silver, gold or
aluminium. These are, of course, highly conductive but normally also rather
opaque.
Figure 3.41 Large glass-coating production line. Note the vacuum pumps on the left (courtesy
of BOC Group).
The opacity of the layer is reduced by two means:
using a very thin layer (e.g. 10 or 30 nm)
using matching layers of oxide, intermediate in optical index between the

glass and the metal layer.
The latter relies on an effect that is well-known in optical antireflective
coating practice. The reflectivity of a transparent surface of refractive index
N can be reduced to zero by applying a layer 1/4 of a wavelength thick
(around in common optical practice) of index The coatings are
applied to the inside of the outermost piece of glass so they are not damaged
by scratching in service. They enable a windscreen to have a low electrical
sheet resistance while still transmitting 80% of visible light.
Coincidentally, the resultant windscreen will also act to restrict solar gain
and similar structures can be used on window glass to control solar gain
without affecting transparency too much in order to limit air-conditioning
load in hot conditions. The completed windscreens must have copper
busbars to enable very high currents and voltages (30 A at 60 V) to be
applied to the windscreen to melt ice and demist rapidly.
Similar processes are used to coat plastics. It is even possible, surprisingly,
to coat plastics with glass-like quartz film as the still relatively new process
from AIRCO shows. A quartz layer from 10 nm to a few tens of nanometres
thick put down from a plasma-excited chemical vapour (such as the silane
derivative tetramethyldisiloxane) or by sputtering will adhere strongly. Such
a layer can decrease a hundredfold the permeability of typical barrier
plastics used for food.
Electron-beam coating and printing. The coating of glass substrates with
polymers is just one example of the very common industrial requirement to
coat sheet materials with polymers. One of the high technology methods of
polymer coating is to use electron-beam cured polymer coating. Electron-
beam coatings are now quite common, for example a typical application for
the process is for applying the multiple coatings used for cardboard fruit
juice and milk containers.
Electron-beam printing and coating processes work as follows. The sheet
of material to be coated is first stripped of its air by a gas knife, then the wet
coating is applied. The wet coating has a chemical formulation chosen to
cross-link or polymerise, forming a solid under irradiation from high-energy
electrons. Electrons are accelerated to 100 keV or so in a vacuum and the
powerful beam then exits via a thin membrane into the air approximately
1 mm above the coating surface. High-energy electrons produce showers of
rays (lower energy electrons which can knock electrons or hydrogens off
polymer molecules and form reactive free radicals), which can then cause
the required cross-linking/polymerisation.
Electron-beam printing always employs nitrogen inerting otherwise, the
energetic (100 keV or more) electrons would cause the production of
copious amounts of ozone. As well as being an environmental hazard, the
ozone would damage the coating.
3.8.7 Gases in double glazing
The objective of a double-glazed window is to combine transparency with
the minimum of heat loss. Typical construction involves two glass sheets,
spaced 5 to 45 mm apart, perhaps with additional thin plastic or glass sheets
between, containing a sealed volume of gas and a dessicant material to
prevent moisture condensation.
Heat transfer across a window occurs via radiation, conduction and
convection. Radiation can be reduced with a layer on the gas side of
the panel. This reduces the radiation of long-wavelength
infrared photons from the warm inner panel to the cold outer panel as
described in section 3.8.6. Although gas mixtures with IR absorbing
properties have been suggested for this, they are not in use.
Conduction and convection can be reduced by using a gas filling other
than air. Basically, switching to a higher molecular weight polyatomic gas
reduces conduction but increases convection because of the lower viscosity
of a polyatomic. The optimum is a heavy monatomic gas, which is superior
even in thick windows. For thinner windows, heavy gases offer substantial
energy saving advantages.
Argon is now in use as a filling gas superior to air, as is for thinner
windows. Krypton is superior again, although very expensive (Table 3.14).
3.8.8 Gases in ceramic firing
Much of the making of clay objects is in essence very simple and apparently
low technology. Natural clay is moulded while wet and plastic into a shape
which is then slowly heated up to convert it into a rigid, durable artefact.
This simplistic description, however, hides a number of subtleties. First,
natural clay only occasionally has the correct properties for a range of
products and must be blended with other ingredients. Typically other
refractory ingredients, often in carefully controlled particle sizes, such as
alumina or silica (sand), are added. Carbonaceous additions that will assist
with firing and oxidation state and porosity may be needed, for example
powdered coal or straw.
Second, heating the clay is a delicate process. First the large amount of
moisture in the wet clay mix must be driven off slowly, to avoid cracking, but
efficiently, the artefact shrinking as this happens. Then, as temperature
increases, the carbonaceous materials must be activated, and perhaps ful l y
oxidised and carried away in gaseous form, i.e. At certain
higher temperatures various minerals undergo phase changes. These points
must be passed slowly to avoid weakening or cracking the clay. The two
forms of silicon dioxide, quartz and cristobalite, interconvert, for example.
The final result is a strong, though brittle, polycrystalline article. Overfiring
must be avoided as fired clay, although very heat resistant, can be distorted
or even melted, depending on its composition, at still higher temperatures.
Colour often depends on the state of the oxidation of included iron
compounds, with green raw clays containing Fe(II) complexes turning to
brick red fired products with Fe(III) coloured minerals. Carbonaceous
inclusions are often added to rough clay products such as bricks, as they
decrease the energy needed for firing but lead to blue or black spots and
core (black heart). Spots may be desirable to some extent in antique
rustic non-structural facing bricks. However, black heart is undesirable in
excess and in structural products as it leads to very weak products. Black
heart problems can be avoided by injecting oxygen during firing. The
Messer Griesheim company in Germany has shown the advantages that
can be gained by judicious injection of oxygen during the firing process.
For many ceramic tiles and bricks (rustic bricks, for example) a dark
colour or dark spots are produced by maintaining an atmosphere with
little oxygen or even slightly reducing conditions. The traditional way of
doing this is to fire bricks in a gigantic clamp. In this method, hundreds
of tonnes of bricks are placed above a bed of coal, insulated with turf on
the top, and the coal ignited. After a month or so, fully fired, typically
rustic style bricks are obtained. The method results in darker colours, and
retains black or blue spots due to carbonaceous inclusions, owing to the
low oxidation potential of the gas retained in the clamp.
This traditional method is very labour intensive, however, and the
brick quality and yield are low. Modern brickworks are now making
better quality bricks of similar appearance via low oxidising atmospheres
in their conventional continous or batch furnaces provided by inert gas
generators (IGGs). IGGs are common in the brick industry for
maintaining an inexpensive inert atmosphere. They generate a very crude
grade of mixture via controlled combustion (see section
2.2.1).
Advanced ceramics production. Some advanced ceramics cannot be fired
in air and require special furnace atmospheres for their preparation.
Advanced ceramics is, of course, a very wide group. However, some of
the most important members of the group are the hardness compounds
WC, SiC, SiAlOH, the magnetic oxide compounds, such as ferrites, and
thermal shock resistant compounds, such as
The removal of binders from the green material is often enhanced by
oxygen addition to the initial low-temperature burnout part of the firing
cycle.
Ferrite and can both be advantageously fired in mixtures of
nitrogen with hydrogen. The presence of a reducing atmosphere is
essential for the latter as at the extremely high sintering temperature,
1800C or more, is oxidised, even by trace oxygen in an inert gas.
Hydrogen sintering of alumina is used to achieve a high density
translucent alumina, which is used in artefacts such as laser discharge
tubes and high-pressure sodium lamp tubes where high strength at high
temperature is required. The high pressure sodium lamp provides high
efficiency in a whitish (actually pink) lamp, so translucence is vital.
3.8.9 Electrical power generation
Hydrogen alternator cooling. The thermal conductivity of hydrogen is very
high indeed (seven times greater than that of nitrogen and 20% or so higher
even than helium) while its viscosity is less than half that of air or helium.
This is important when thermal transfer is considered vital but a liquid
coolant cannot be employed. Very large (hundred megawatt or more)
electrical generators are just such a situation and many of these are now
cooled by circulating hydrogen around the windings and having a water-
cooled heat exchanger. The very high performance given by hydrogen is the
reason for its continued usage. In less critical applications such as wafer
backside cooling in semiconductors (section 3.5.7) the nonflammable nature
of helium makes it the gas of choice. In alternators, however, potential
safety problems are overcome by stringent precautions.
Sulphur hexafluoride high-voltage circuit breakers. Sulphur hexafluoride is
an excellent electrical insulator, giving dielectric strengths. Unlike oil it is
also nonflammable. It is therefore used in high-voltage apparatus where a
gaseous insulator is desired but where the maximum voltage should be
endured before electrical breakdown occurs. This situation occurs in Van de
Graaff accelerators of the type used in nuclear physics work. These need a
few megavolts d.c. potential differences in a compact laboratory-sized
apparatus to accelerate nuclear particles to megaelectronvolt energies. A
much more common application, however, is the application to high-voltage
electrical circuit breakers.
Electrical arc discharges are initiated by the rapid acceleration of a few
free electrons to energies at which they can collide with atoms and ionise
them. The additional electrons released are in turn accelerated and an
avalanche is built up within microseconds. The halogen atoms of tend to
mop up any free electrons, preventing the initiating action, and will not
easily release electrons in collisions, so preventing avalanche build-up and
arc formation.
When a mechanical circuit breaker operates on a high-voltage and
high-current electric circuit, current is driven by the inductance of the
upstream and downstream circuit to carry on flowing. A high-voltage will
build up across the gap and an arc will tend to occur. If allowed to carry on
more than momentarily, this arc would destroy the circuit breaker, so a
number of techniques have been evolved to extinguish the arc. Breakers can
simply use a very large air gap. The air gap, and the overall size of the
apparatus, needed is grossly large, however, for 132 kV or higher voltages
Figure 3.42 High-voltage rotating arc circuit breaker. Current from inputs (left) is led via
the moving contact bars to the outputs (right). The circular coils visible in the lower photograph
are the rotating arc devices. The mechanical operating rod and a small sack of moisture
absorber are visible at the bottom (courtesy of Hawker Siddeley Switchgear).
and high current. Air-blast circuit breakers reduce the size needed by using a
forced blast of air to extinguish the arc. A high voltage circuit breaker using
is also considerably smaller than a similar air breaker unit because of the
higher electric field that can be tolerated. breakers can also use a flow of
gas but the lower powered gas flow needed has led to the term puffer to
describe the gas arc extinguishing mechanism. Puffer breakers are used on
100 kV or more tranmission lines and are typically contained inside long
cylindrical ceramic insulator pressure chambers containing the The
mechanism is elongated to fit inside the insulator chamber and operated by
means of a long insulating rod that has a sliding seal in the base of the
chamber.
An ingenious use of the dynamo principle is made in rotating arc
switchgear (Figure 3.42). In this, the arc is transferred to a circular contact
connected to a solenoid coil. The arc current itself then creates a magnetic
field which causes the arc to move around the circular contact surface: the
cool surfaces and cool gas, together with the free electron-absorbing
action of the gas, extinguish the arc rapidly.
Pressures used vary from 1 bara up to 5 or 6 barg, the gas being typically
contained in a metal cylindrical pot pressure chamber. Many designs have
only one or two rotating seals penetrating the pot, in order to minimise
leakage. The gas is slowly decomposed by the arcing process. Metal
fluorides and similar are formed, which are often absorbed, along with
moisture, by a cartridge or sack of absorber often molecular sieve. Smaller
units using can be sealed for life (10 years or so) and slight losses of gas
pressure (say 1% per annum) are acceptable.
Other applications in electricity. There are many and varied small appli-
cations for gases in the electricity industry. Nitrogen is often used, for
example, along with high-grade oils, to insulate underground high-voltage
cables. Nuclear power plants have many small applications for gases.
Ultrahigh purity (argon free) oxygen is used in nuclear applications
because Ar-40 is activated by neutron absorption to Ar-41, which is
radioactive with a half-life of 1.8 h. This would constitute a distinct danger in
the event of gas leaking from a gas-cooled reactor. Early low-power reactors
used air for cooling and discharged Ar-41 into the atmosphere. This is no
longer acceptable and in any case reactors are now much more powerful and
require a forced gas circulation system if they use gas cooling at all (most use
water). Somewhat higher power early reactors used nitrogen or helium for
cooling; the latter, being especially satisfactory as it required lower power to
pump for the same cooling effect, has a low cross-section for the absorption
of neutrons and is also a neutron moderator. Carbon dioxide is used in most
British reactors and to a lesser extent elsewhere, although some of the
reactor power is wasted in pumping the coolant gas compared to water-
cooled reactors. Carbon dioxide has reasonably good heat-transfer proper-
ties and has a low cross-section for the absorption of neutrons.
3.8.10 Gases in lighting
Incandescent filament light bulbs. The filament lamp is a simple concept.
Inside the glass envelope is a thin filament of tungsten which glows white
hot, emitting light, when it is heated by a relatively high current. The first
lamps were very straightforward, simply evacuating the glass bulb around
the filament (then of carbon). There is, however, a fundamental com-
promise in incandescent lamps, i.e. the trade-off between efficiency and
lifetime.
The problem of a filament lamp is as follows. The higher the temperature,
the higher the efficiency and better (whiter) the light colour. However, as
the temperature is raised, the filament evaporation rate rises. Filament
evaporation leads to two problems. First, blackening of the envelope as a
result of a deposit of tungsten, with consequent loss of light and, second,
thinning of the filament, leading to failure at hot spots. A little observation
will confirm that failed bulbs are often blackened, indicating how much
tungsten has been deposited from the filament, whilst bulbs just before
failure often seem to be brighter and whiter than usual because the thinner
filament runs hotter.
Hot spot failure occurs as follows. The filament is not perfectly uniform in
cross-section. In thinner places, the filament will become slightly hotter than
average when current passes. These hot spots soon become worse in a
vicious circle if the bulb is run a little above its rated current. The hot spot
loses filament by evaporation and the filament becomes slightly thinner still.
This leads to a further increase in temperature, which leads to more
evaporation and so on, until the filament fails at the hot spot.
For small lamps, this is all there is to it. Bulbs of a few watts or less are
evacuated to a good standard with an oxygen/moisture absorber, the
getter.
For larger lamps, however, there is another parameter that can be usefully
played to improve performance, the gas filling. Argon reduces evaporative
loss of filament, though at the cost of increasing convective/conductive heat
losses (these rule out gas filling at low wattages). The lower evaporative loss
leads to a longer lifetime, or enables a higher temperature to be used, giving
a greater efficiency.
For filled bulbs, the higher the pressure of inert gas, the lower the
evaporative loss from the filament. This is not a simple effect of gas pressure.
At the moderate pressures used (from a few torr to a few bar), Daltons law
of partial pressures would hold approximately, meaning that the presence or
absence of argon would not affect the behaviour of the tungsten in
evaporating directly. Probably the effect is that a gas filling of any sort simply
forces the tungsten evaporating to form an equilibrium gas, with tungsten
atoms both being evaporated and condensed, around the filament. At lower
vacuum pressures tungsten atoms can simply travel radially and stick to the
glass envelope directly. The cooling effect of the gas may also play some
part. Hot spots may be suppressed by the conductive/convective heat
transfer of the gas.
In higher voltage (e.g. mains 240 V) bulbs, pure argon could lead to
arcing, depending on the electrode configuration, so these bulbs usually
include 10% nitrogen in the gas to ensure arcing will not occur. Krypton in
tungsten filament bulbs provides increased efficiency because of lower
thermal losses; the heavier krypton gas has a lower thermal conductivity.
Quartzhalogen bulbs. Quartzhalogen bulbs are, as the name suggests,
filament lamps with a quartz envelope and a halogen in the mostly argon gas
filling. It is often thought that they can operate at a higher temperature and
at a higher efficiency because of a self-healing action on hot spots induced by
the halogen but this is incorrect. The halogen, typically iodine, reacts with
the evaporating tungsten atoms, preventing them from being deposited on
the bulb wall by reacting with them to form iodide. The tungsten iodide
molecules diffuse back towards the filament where they decompose, giving
up their tungsten back to the hot filament and releasing iodine for further
reaction. This regenerative action means that no tungsten is lost from the
filament, although hot spots are not prevented because redeposition occurs
all along the filament or even preferentially on cool spots. The quartz
envelope is necessary to handle the higher temperatures required (250C)
which prevent deposition of tungsten iodide on the bulb.
It is the fact that tungsten is not deposited on the envelope that gives the
true advantage of the quartz-halogen lamp. A smaller lamp can be used (as
small as 1% of the volume of a conventional lamp), which can then be run at
higher gas pressure, further suppressing tungsten loss from the filament and
giving longer life and/or greater efficiency. A small lamp also has little
convective heat loss.
The quartz-halogen bulbs in use today do not use self-healing but do
obtain better performance by the use of halogen. Bulbs with a true
self-healing action are possible, however, using fluorine. With fluorine in
the bulb, the whole filament reacts at operating temperature, to a small
extent, to produce tungsten fluoride. This compound decomposes at very
high temperatures to tungsten, decomposing more rapidly in the hotter
spots. The hot spots thus become thicker until they reach the same diameter
as the rest of the filament, when equilibrium is reached. There are, however,
severe materials problems attached to using fluorine inside a lamp. Fluorine
attacks both the glass or envelope and the filament supports. Hence there
are no mass-produced lamps of this sort available.
Electric discharge, fluorescent and neon lamps. The rare gases, with the
exception of argon, were discovered by their optical emission characteristics
in an electric discharge and these characteristics are also responsible for one
of their main uses. Argon at a few torr pressure, is used in many fluorescent
lamps as a gas filling to allow convenient starting. The common fluorescent
lighting tube is in fact a mercury vapour tube with much of its light output in
the ultraviolet. The ultraviolet emitted is absorbed by the phosphor coating
on the inside of the glass and re-emitted at visible wavelengths. A mixture of
phosphors emitting different colours ensures that the light finally emitted is
approximately white. The argon filling enables a small initial discharge while
the heaters warm up the mercury in the lamp ends. Once the mercury is
vaporised a little, the light from the low-pressure argon discharge is
swamped by the powerful mercury emission.
The argon and a little mercury vapour together form a Penning mixture, in
which electrons in the discharge can raise the the majority of Ar atoms to an
excited state by collision. The excited Ar atoms can then ionise the Hg atoms
in a low-speed collision because the excitation energy of the Ar (11.6 eV) is
very similar to, but a little larger than, the ionisation energy of the Hg
Figure 3.43 Low-cost lamp-starter circuit using Kr-85 tube.
(10.4 eV). This process increases the degree of ionisation induced and
reduces the mi ni mum voltage needed to set up a discharge (the striking
voltage).
Other common discharge lamps use sodium instead of mercury, giving a
bright yellow light, and these are very efficient. High-pressure sodium lamps
are also now used, which give a pink light, although severe materials
problems had to be overcome in containing high-pressure hot sodium
vapour before these became practicable. The discovery that a mixture of
metals could be used in a discharge lamp by including a halogen led to the
metal-halide lamps. These give a better colour of light than sodium whilst
maintaining a high efficiency.
Both sodium- and mercury-based fluorescent lamps require special
starting conditions. Heating filaments at each end to warm up the sodium
and mercury need to be supplied, along with a striking voltage. In some
modern equipment, start-up is controlled by an electronic control system. In
older and simpler systems, a gas discharge starter tube may be employed.
This includes a minute amount of radioactive gas, either Kr-85 or H-3
(tritium) and a bimetallic switch (Figure 3.43). The starter tube initially has
the full mains voltage across it and, with the help of ionisation from the
radioactive gas, immediately strikes up a discharge and rapidly warms up the
bimetallic switch, which closes, shorting out the starter itself but warming
the filaments connected to it. After a few seconds, the bimetallic contacts
cool and cut off filament current, allowing the main fluorescent tube to start.
Once the latter is running, the voltage available is reduced by the series
inductor and is insufficient to close the starter tube contacts.
Other lamps rely wholly on the optical emission of their gas, rather than
metal ions. Xenon is an example. It can be used in a high-current pinched
discharge to give a bright white light. Another purely gaseous emission lamp
is the neon lamp. This actually uses another Penning mixture, i.e. neon with
a little argon. Neon discharge tubes are familiar as the glass tubular emitters
which can be bent into extraordinary shapes and are used in illuminated
signs. The red tubes do actually use neon gas at a few torr pressure but
careful selection of different gas/vapour mixtures, often including neon, can
give alternative colours. To judge by the night-time lights in neonophile
cities such as Tokyo, most colours in the rainbow seem to be available!
Gas lasers. Gas lasers rely on an electrical discharge causing electron/ion/
molecule collisions resulting in excited energy states. In this way, with the
correct choice of gas mixture and pressure, a population inversion (high
energy states more numerous than low energy ones) can be built up. Once
the population inversion is strong enough, a single stimulating photon can
result in a cascade of photoemission events. All the stimulated emission
photons are precisely in phase with the stimulating beam. With reflection at
each end of the discharge, the photons can be built up into a strong coherent
beam, part of which can be released from the tube by a hole or
semitransparent part in the end mirrors.
Carbon dioxide lasers actually use a subatmospheric pressure gas mixture
of (1040%), (510%) (and sometimes CO) with balance He,
although it is the molecule which emits the laser light. These lasers emit
at 10.5 in the far infrared and can be made exceptionally high in power
output; several kilowatts is quite normal. Such high-power beams can easily
melt most solids, even in quite thick sections. For this reason, these lasers
are the preeminent metal working laser (section 3.2.6).
lasers developed fairly early in laser development and rapidly grew in
power and efficiency, as knowhow grew, until today hundred kilowatt, 15%
efficient machines are available; the most efficient and powerful lasers in
regular use. The key discoveries in their development were as follows.
The discovery that nitrogen vastly increased the power output of a

laser discharge. This happens because excited nitrogen molecules can
transfer their energy efficiently to ground-state molecules by
collision and produce the required population inversion in the energy
states.
The discovery that addition of helium offers additional power increases.

These are thought to be due to Hes ability to cool the molecules
more rapidly back to the ground state from which they are excited by
into their lasing levels, and because He aids the electrical discharge
process.
Improved techniques for cooling the whole laser system, both gas and
tube. Many new designs for lasers focus on improving the cooling,
typically by circulating the gas mix more rapidly.
With these developments, the laser has become powerful and
efficient enough to be a useful tool for industrial cutting of all materials,
from dress fabric to steel armour plate, and in the welding of metals.
Other gaseous lasers are possible using similar principles to that of the
laser. Excited helium can transfer energy to neon atoms, which can
then give a laser action. He/Ne lasers rely on an electrode gas discharge in a
mixture of about 10% Ne in He. They are the cheapest lasers available,
costing only a few tens of dollars for a 2 mW device. They emit red light at
623 nm at up to a few tens of milliwatts. They are often used as a highly
convenient source of a narrow light beam, such as in supermarket checkout
scanners. These sorts of applications ignore many of the extraordinary
qualities of lasers such as monochromaticity and coherence, but are far more
important commercially than applications, such as making holograms,
which use those qualities.
Other common gaseous lasers are the argon ion systems and the excimer
laser. Argon ion lasers rely on transition between two levels in the ion
and emit in the green at 440530 nm. They are commonly used when a
powerful, continuous (tens of watts) visible beam is needed. A similar
krypton laser can be made with outputs at different frequencies from 460 to
690 nm, and an almost white laser by mixing these gases in the laser.
Krypton fluoride and similar excimer lasers rely on transitions in excimers,
i.e. pseudo-molecules formed between the first excited state of a noble gas
atom (which in this state is as reactive as an alkali metal atom) and a halogen.
They emit in the deep UV region at 200 nm and are currently being used
mainly in the laboratory, although industrial applications such as photoli-
thography of submicron silicon integrated circuits are now beginning.
Excimer lasers use a mixture of 5% fluorine or HCl, which is then blended
with neon, krypton or xenon for use in the laser.
3.8.11 Balloons and airships
Although many gases are lighter than air, only three are common in balloons
and airships: hot air, hydrogen and helium.
All aerostatic craft work by displacing a greater weight of air than their
own weight. This is not an easy goal to achieve, as air only weighs
approximately Leonardo da Vinci drew a picture of a potential
balloon craft based on evacuated copper globes. This is unworkable,
however, owing to the impossibility of making the globe simultaneously
strong enough to resist collapse due to external air pressure and lighter than
the air displaced. The only practical solution is to fill the envelope of the craft
with a gas lighter than air. Fortunately, weighs only and He only
so their performance is close to that of da Vincis vacuum.
The envelope of a balloon, even that of a hot-air balloon, must be selected
carefully to minimise gas losses. Hydrogen and helium are both small
molecules and diffuse rapidly through most materials. The most obvious
flexible materials, rubbers, plastics, and textiles, all have problems. Textiles
can be used for hot-air balloons, but only after a plastic proofing compound
has eliminated most of their porosity. Elastic rubbers, although convenient,
because a rubber envelope can expand and contract easily with pressure
changes and packs away more easily, provide little protection against gas
Figure 3.44 Meteorological balloon using hydrogen as a lift gas. The rubber balloon supports a
large wire mesh corner reflector from which is suspended a small instrument pack (courtesy of
the British Meteorological Office).
diffusion. Low-cost plastics such as polythene are also highly permeable.
The least expensive barrier plastics are nylon film and PVDC-coated types.
Even better than this is to use a metal film as part of the plastic film
composition, metals providing an almost perfect gas barrier.
Toy helium balloons may sound like a frivolous use of gas but they are in
fact a business worth hundreds of millions of pounds worldwide. Helium
diffuses out of rubber balloons in less than a day so fancy decorative balloons
are normally made out of aluminised which will stay filled for a
week or so.
Meteorological balloons. For applications in which fire risk is not a
consideration, the use of hydrogen offers useful cost savings over helium.
Weather balloons are such an application. Balloons can be used to track air
currents and relay temperature, pressure and humidity high in the atmos-
phere back to meteorologists. Despite advances in remote imaging from
space and chains of interlinked storm radar units, it is still useful to be able to
probe the atmosphere directly with balloons.
Meteorological balloons vary in size from less than to hundreds of
cubic metres (Figure 3.44). The smallest are used to carry simple radar
reflectors (corner reflectors similar to those seen on yacht masts), whilst the
most sophisticated balloons are used by researchers to study the atmosphere
right up to the edge of space. Huge numbers of simple radar reflector units
are used. They rise in the air, the rubber envelope expanding as they gain in
altitude until the rubber is fully stretched. If the balloon still has enough
buoyancy, however, it carries on rising until its internal pressure causes it to
burst.
65
With a plastic envelope that is relatively resistant to stretching, balloons
can be made to float aerostatically at a more or less predictable height, i.e.
the height at which their density with the envelope fully dilated equals the
surrounding air. Very high altitude balloons use a simple thin, lightweight
PE or similar envelope, which is filled with only a small volume of gas at
ground level. In this way, balloons can easily be made to ascend to over
30 000m (100 000 ft).
Methane and anhydrous ammonia, despite having a lifting effect only half
that of hydrogen, are occasionally used for filling meteorological balloons.
Methane, in the form of piped mains gas, is the cheapest filling. Ammonia
has the advantage that it can be transported efficiently in lightweight
cylinders in liquefied form at only 6 bar pressure. This means that a similar
lifting effect to a cylinder of hydrogen can be obtained in a package several
times lighter in weight; an advantage for balloons launched, for example, by
expeditions. However, ammonia is both flammable and toxic, and contact
with moisture must be avoided as the ammonia would dissolve, reducing the
balloon volume.
Airships. Ever since the Hindenberg airship fire, hydrogen has been
regarded as exceptionally dangerous. Yet it is interesting to note that in their
30 years of operation (roughly from 1910 to 1937) airships achieved many
millions of passenger-miles, many of them on very long routes such as that of
the Graf Zeppelin from Germany to South America, with a safety record
better than the aeroplanes of the day. The Hindenberg incident, which may
have been sabotage (Mooney, 1973) killed 13 passengers but 23 escaped, as
did two-thirds of the crew. This would be a rare occurrence in an aeroplane
crash.
The airships seen occasionally above big cities and above sporting events
are helium-filled. The helium is a considerable cost. For each tonne of
payload, around of He is needed, which, if a price of $5 per is
assumed, means that their gas filling costs at least $5000 per tonne. It also
needs constant replenishment as losses due to leaks and diffusion must be
replaced. These small airships are semi-rigid; their shape is maintained by
internal pressure rather than the rigid frame of the earlier Zeppelins.
65
The lightweight mesh reflectors and instrument packages are occasionally seen descending,
apparently from nowhere, after this has happened.
Figure 3.45 High-capacity (13 MW) triple nozzle burner for use in hot-air balloons. The coils of
tubing are for rapid vaporisation of the liquid propane fuel (courtesy of Cameron Balloons
Ltd).
Although a few passenger-carrying machines are maintained, the main
applications for these are for providing a steady aerial platform for television
cameras and to present a slow-moving aerial signboard for advertising. In
addition to large-scale graphics, some have giant LED display panels
covering large areas of the envelope for nocturnal use.
Hot-air balloons. Hot air may seem a curious choice for a balloon gas as it
would appear to be very inefficient, because it is only very slightly less dense
than cold air. However, if the air inside the envelope can be maintained at a
mean 50C above a 20C (290 K) ambient, then the ideal gas laws can be used
to show that the lifting capacity is
which is around six or seven times less than that of hydrogen or helium. This
may sound disastrously less efficient until it is remembered that this means
that the balloon diameter must be increased by a factor of the cubic root of
six or seven, i.e. less than a factor of two bigger.
The reason for the current popularity of hot-air balloons for sporting
purposes is the ease of use and low cost of operation of such an aerostat.
After use the balloon may be packed away each day without loss of a
valuable gas, whilst a relatively simple but very powerful propane burner
provides a readily adjustable means of controlling the balloons lifting
capacity (Figure 3.45).
Modern hot-air balloons rely on a certain amount of up-to-date technology
for their function. Although the wicker basket suspended beneath them is
old-fashioned, most of the rest is not. They use lightweight aluminium or
stainless steel propane cylinders, the gas being burnt in ultrahigh capacity
burners with vaporiser coils heated by the flame itself. The envelope
construction typically includes fireproof cloth in the lowest part and
ultrastrong suspension tapes, cords and envelope fabric.
A combination of hot air and helium has occasionally been used in balloons.
A lifting balloon of helium is supplemented by hot air; the former provides
most of the basic lift, the latter provides controllable additional lift. Such
balloons have been used for altitude and distance record-breaking, and for
providing a stable and long-lived high-altitude platform for scientific work.
3.8.12 Aerosol spray cans
An aerosol spray can in its basic form today comprises an aluminium or
tinplate can filled with a mixture of propellant, typically 50 to 95%, and a
smaller amount of the active product dissolved in that propellant. A dip-tube
from the bottom of the can leads to a valve, which when depressed releases a
high-pressure stream of product via a spray nozzle, driven out by the gas
pressure in the ullage space. Until the late 1980s, the propellants generally
favoured were CFCs and these are still used in medical products. With that
exception, products today use flammable hydrocarbon mixtures, mainly of
isobutane and propane.
A variant of the standard aerosol can is the bag-in-can system in which the
product is contained in a flexible bag, pressurised by a propellant below the
bag. These cans avoid the need for the propellant to be dissolved in or mixed
with the product. This is particularly difficult when oil and water must be
mixed, which is what is attempted with hydrocarbon propellant and
water-based product. Bag-in-can aerosols also enable corrosive materials to
be dispensed because the product never touches the can. They are most
commonly used for dispensing gels
66
or other products where the presence of
propellant in the product is undesirable. Bag-in-can aerosols have had to
overcome a number of technical problems, principally permeation of
propellant through the bag and poor spray characteristics because of a lack of
propellant to flash-off from the product. A selection of aerosol can types is
66
Bag-in-can aerosols can also be used for shaving gels, where an additional subtlety is added.
These gels are made with a small addition of isopentane, which happens to boil at a little under
body temperature. The dispensed gel is clear when emitted, therefore, but rapidly foams up
when it contacts warm skin prior to shaving. Although dismissed by some, these shaving gels are
now popular.
Figure 3.46 A selection of aerosol can types.
shown in Figure 3.46: propellant can be added via the bottom plug or via the
valve.
The permeation problem can be solved by using gas-proof coatings such as
Al foil on the flexible bag. The spray characteristics can be made acceptable
by the use of higher (up to 130150 psig) pressures and a mechanical
break-up nozzle (MBU). The MBU nozzle works by swirling the product in
the duct behind the nozzle. As the product is forced through the tiny orifice,
angular momentum is conserved, causing the fluid to accelerate its rotation.
On exiting the nozzle, a fine spray is formed. MBU nozzles are limited in
flow rate so to achieve small particle size (say 60 on a water-viscosity
product) a low flow rate MBU nozzle is needed. An acceptable compromise,
e.g. 80 particles and can be obtained.
An aerosol spray can is a convenient means of dispensing small amounts
of a liquid or fine powder product from a hermetically sealed container. It
protects the product being sprayed from degradation by the air, which is
normally seen in bottles and similar, and enables the use of smaller
quantities of product than might be the case with brushing, for example.
With an aerosol can there is no need for solvents to clean brushes. The rapid
shut-off of pressure when an aerosol can nozzle is released means that there
is little spitting from an aerosol can. Pump spray guns have a problem with
spitting as well as having a coarser and less controllable spray particle size.
It is largely as a result of these genuine technical advantages that aerosol
packages have grown rapidly from early crude devices for applying
insecticide in the 1940s to the present plethora of aerosol cans. The aerosol
can has, however, accumulated a bad public image over the years. The first
blows against the aerosol undoubtedly came from the aerosol paint can. The
many plain concrete surfaces of 1960s and 1970s architecture were ideal
substrates for impromptu artwork and slogans, and the aerosol can provided
a clean, neat, pocketable package and a high speed of delivery for the
unauthorised artist.
More serious objections came in the 1970s and 1980s with the growing
realisation that the stratospheric ozone layer was thinning and that CFCs
might be primarily responsible for ozone loss (section 2.5.12). There is a range
of CFCs and mixtures of CFCs with room temperature vapour pressures of
between 15 and 65 PSIG approximately, and virtually all of these have been
used. Probably the most popular were CFC-11 and CFC-22. Greener CFCs,
such as HCFCs, are now becoming available but are too expensive for a
throw-away product, probably have some ozone-destruction potential and
certainly would contribute to global warming if adopted on a large scale.
The aerosol industry was persuaded by pressure from environmentalists to
move to hydrocarbon propellants. These are basically mixtures of isobutane,
butane and propane with small amounts of ethane and other light
hydrocarbons dissolved in them. These led to some problems, being less
miscible with products than the CFCs, but the addition of suitable co-solvents
usually solved these problems. Yet more recently, environmental lobbyists,
already looking closely at aerosols, have insisted that spraying a considerable
amount of hydrocarbon into the atmosphere, as current spray cans do, is also
undesirable as the hydrocarbons contribute to VOC pollution. Ironically,
hydrocarbon propellants actually encourage the formation of ozone but in the
lower atmosphere, where it is toxic and undesirable.
There are occasional accidents caused by the flammable nature of the
hydrocarbon propellants now used.
67
Fogger-type spray cans, for example,
have occasionally led to devastating detonations. Foggers are typically
designed to dispense their entire contents of paraffin-based insecticide
product. In accidents the paraffin is aerosolised and then somehow ignited
whilst the air/droplet/vapour concentration ratios are near stoichiometric
values. However, arguably the most serious problem with hydrocarbon
propellants is their abuse by children as an inhaled drug; an unfortunate
practice which is now widespread with frequent fatalities.
Alternative aerosol can systems. There are alternative aerosol packages
that do not need a hydrocarbon propellant. A simple replacement propo-
sition is to use an oxygenated solvent with high vapour pressure. Oxygen-
ated solvents still have a flammability problem but are miscible with more
products, are not responsible for tropospheric ozone production or strato-
spheric ozone destruction, and are less toxic when abused as an inhaled
drug. However, most oxygenated compounds have too low a vapour
67
The author assisted a specialist fire test company to conduct tests simulating the effects of a
can which is accidentally or deliberately overheated near a source of ignition. On these
calibrated barbeque tests some relatively small spray cans produced a violent explosion and a
large (3 m diameter), though t ransi ent , fireball.
pressure and the only one currently in use is dimethyl ether (DME). DME is
today used in many low-volume products such as perfumes but has yet to
advance into higher volumes on cost grounds. Perhaps the building of
large-scale DME production plants (DME is an unleaded petrol octane
improver) currently under way will lower prices and make DME more
attractive.
Simple pressure can also be used. However, in the headspace has a
poor technical performance. There is a large pressure droop when the can
is more than two-thirds used up and to counteract this it is necessary to use a
large ullage space and higher pressure, thicker-walled cans. The resultant
half-filled cans give a poor value marketing image and the thick-walled cans
are expensive. Even with these precautions, cans can degener-
ate to give not a spray but a useless streaming flow; a disaster with a product
such as hairspray or paint. Patented systems have been developed, but not
so far widely adopted, which employ a pressure regulator inside the aerosol
actuator button. This has the advantage of giving a more constant pressure
at the nozzle, which can be designed to give optimum atomisation at the
lowest pressures of the can.
Carbon dioxide and nitrous oxide both dissolve to a much greater extent
in some products and can be used in a simple aerosol can to give a reduced
pressure droop relative to because of the reservoir of gas which remains
in the product. has always been effective for certain aerosols, for
example engine starter and deicer products, which need pressure at very low
temperature. Liquid under pressure, because it is a good solvent for
many organic substances, could be very useful as an aerosol propellant.
However, the pressure required at room temperature is too high (over 70
bar must be allowed for) and means that this is not possible with tinplate or
thin-wall aluminium cans.
68
is favoured for the dispensing of food products. It does not add an
acidic flavour, as would be the case for but does foam the product to a
controllable extent. Filling machines must all be carefully sterilised. Cream
is the product most often packaged. The dissolved expands the liquid
cream into a light whipped cream consistency foam on exiting the nozzle,
thus conveniently avoiding the need to whip the cream mechanically.
A system called Polygas, currently being developed for using uses a
flexible bag inside the can pressurised by means of absorbed into a
solvent/solid absorber mixture in the propellant space. It has the advantage
of a pressure droop lower than nitrogen but is operable at higher pressures
than hydrocarbons. (The latter cannot be used above about 70 psi because
aerosol cans must withstand heating to 55C, at which temperature
68
The BOC company offers an aerosol system of this type using conventional high-pressure
gas cylinders as the containers for its system of dispensing smells and for some of
its pesticide formulations.
hydrocarbons of above 70 psi room temperature vapour pressure would be
in danger of splitting the can.)
Problems with aerosols. There are clearly potential materials problems in
aerosols. Typical aerosol products contain from two to twelve ingredients,
whilst cans employ deterioration-prone tinplate, plastic and rubber com-
ponents. Some of the precision valve or button parts can fail if they swell
slightly in a solvent additive. Metal corrosion is also a problem; alkaline
products are unsafe in aluminium cans and acidic or chlorided products are
unsafe in steel cans. Rubber seal failure is not uncommon with new
formulations or propellants. Acetone and DME, for example, are powerful
solvents, causing strong swelling or shrinking reactions with most cheap
rubber seal compounds.
Permeation through plastics is always a potential problem, as mentioned
above, for bag-in-can systems. As well as propellant loss, the product can
diffuse through cheap, thin single-layer bags into the propellant space. If a
bag-in-can system was chosen to protect the metal can from a corrosive
product in the first place, this could be serious.
A small percentage of aerosol cans will leak propellant. This is disappoint-
ing to the consumer who will be unable to use the product. To avoid this, and
to provide a modicum of safety against cans which might explode because of
faulty manufacture, cans are legally required to be water-bathed. The
water bath is a large tank at 50 or 55C in which cans are heated for a few
minutes. Small leaks cannot be found in a water bath but gross leaks will
emit a tell-tale trail of bubbles. Weak or over-gassed cans will split.
3.8.13 UV coatings
Most coatings, paints, inks and varnishes, are still today applied in wet form
with the resins that will form the final coating dissolved in a liquid, very often
an organic solvent. Wet coatings using water as a solvent are tricky to make
waterproof and are slow to dry, hence the attraction of solvent-based
formulations. However, there is a drive towards low-solvent emission with
regulations on VOC emission becoming steadily more stringent throughout
the world. This drive means that other means must be found to form
coatings. Hot-melt dipping and dry-film lamination are limited in appli-
cation. Thus wet coatings must be devised that dry without the aid of
solvents. The oxygen-setting resins, such as the time-honoured linseed oil,
will no doubt have some niche, along with the catalytically set epoxy
coatings. However, wet coatings set by ultraviolet light are growing rapidly
and will continue to do so.
UV coatings are formulated from monomers, additives, such as ben-
zophenones, and polymers dissolved in the liquid monomer. UV coatings
set by polymerising the liquid monomer with active radicals. The latter are
formed when the UV photons strike sensitive molecules, photoinitiators,
such as benzophenones or thioxanthones, and split them into two active free
radicals, which then start a chain reaction of polymerisation, acting as
catalysts. Gas-inerting with nitrogen avoids undesirable side reactions,
allowing the use of smaller amounts of photoinitiator and other expensive
additives. UV photons have much less energy than the showers of electrons
produced in the electron-beam system and ozone production is small
compared to electron-beam. However, the oxygen of the air, together with
the small amounts of ozone, cause undesirable chain-terminating reactions,
leaving the coating sticky.
UV curing is currently used for around 510% of the volume of inks and
varnishes used in the printing industry. Printers, taken to include all
thin-sheet manufacturers (e.g. floor tiles and wallpaper, not just docu-
ments), are the largest users of coatings, with wooden products, consumer
hardware and car producers not far behind. Total radiation-cured coatings
now amount to 2% of the entire mass-production coatings industry and are
rapidly expanding by 20% or more per annum. By the end of the century,
8% or so of industrial coatings will be radiation-cured. Of the radiation-
cured systems, the most common, over 90%, are UV cured; the remainder
use an electron beam for curing. The prospects for the UV curing industry as
a whole are exceedingly good. However, it is unclear whether gas-inerted
UV coatings will also grow in usage.
Ten years ago, when UV curing was still in its infancy, many laboratories
had test rigs using inerting of the curing unit. However, the high cost of
put them off scaling these units up. Nitrogen improves quality and can
reduce costs but is not essential. Early in the UV coatings industry there
were problems with gas inerting. Gas prices were high and volumes used
were high enough, together with high coating material costs, to put most
printers off. High on the wish-lists of UV printers at the time was the
elimination of the requirement for nitrogen. This was achieved by the
addition of large amounts of photoinitiator and other additives.
Gas inerting could offer the following advantages, however:
elimination of the need for large amounts of coinitiators like amines,

which increase costs and degrade varnish quality and solvent/water
resistance;
faster cures, enabling faster press speeds;
lower surface tackiness (tackiness is caused by oxygen inhibiting cure at

the surface);
lower UV lamp power needed;
lower quantities of photoinitiators needed;
elimination of undesirable ozone production.

If a package of optimised coating/gas supply/UV lamps with guaranteed
performance and appreciably lower cost were available the system might be
widely adopted. At the cost of increasing photoinitiator or coinitiator levels
slightly, it may be possible to employ PSA quality gas in UV cure inerting,
particularly if PSA gas purity improves. This would give another supply
route option but would only be useful if it enabled reduced overall costs
relative to liquid nitrogen. Some new formulations are being tried out which
may perform satisfactorily on 99.5% or 99.8% purity nitrogen, matching
low-cost PSA nitrogen.
3.9 Medicine and science
3.9.1 Gases for medicine
Some of the earliest applications of gases were to medicine. This may have
been because of the tendency of early experimenters, after a few cautious
preliminary sniffs, to inhale deep breaths of a new gas to see what happened.
Just like todays glue-sniffers and solvent-abusers, the early pioneers
sometimes came unstuck doing this. However, the result of this hazardous
early work was the rapid development of the medical possibilities of oxygen
for improving breathing, and nitrous oxide and others for anaesthesia.
Gases in breathing. Breathing is the process of exchanging produced
in the body for that is needed to convert food into energy. The human
lungs have about the area of a tennis court, most of this is the enormous area
of the alveoli, the tiny terminal branches of the tree structure of the lung
tubes. In the tiny delicate alveoli, oxygen from the air can diffuse through a
thin membrane into the blood where it is absorbed by haemoglobin complex
and carried off to oxygenate the rest of the body. A high-speed mass
spectrometer sampling the air at the end of the nose will see the carbon
dioxide and oxygen concentrations varying.
dissociation is the key to blood function; not only does
haemoglobin absorb easily, it also gives it up easily, being somewhere
close to physical absorption in its reversibility. The solubility of oxygen in
the blood is very high; within the chemistry laboratory its performance can
only be approached by exotic perfluorinated compounds.
69
Although oxygenation is its most difficult technical feat, blood also has a
69
The latter have been proposed and developed, but not yet accepted, as universal blood
substitutes for emergencies. Small creatures, such as mice, can survive for extended periods
with fluorocarbon in their lungs instead of air. The first application for these fluorocarbons in
humans may also be in lungs rather than in blood, in the treatment of premature babies. A
low-viscosity perfluorinated compound is used to treat premature babies with insufficiently
formed lungs. Instead of oxygen gas therapy, the fluorocarbon liquid, saturated with oxygen, is
used to inflate the infants lungs and provide oxygenation for a few days unt i l the lungs have
matured enough to allow gas breathing.
function with respect to carbon dioxide gas. Not only must be provided to
the body but must be removed. It turns out that fairly precise
control is needed. Hyperventilation leads to alkalosis, i.e. is washed
out excessively and the body pH rises. Conversely, hypoventilation leads to
acidosis, i.e. a fall in pH. Under normal circumstances, animals detect
variations and as a result the presence of excessive is a stimulant for
breathing. (However, this works only up to a small percentage in air; beyond
this is toxic and narcosis is possible.)
Oxygen and breathing. One of the first applications for oxygen gas was its
use in reviving people with breathing difficulties. Although important, in
fact only small volumes of oxygen are used in medicine; requirements for
the human body are about 2 kg per day. The reason why such large gas
cylinders seem to be necessary for the maintenance of life is because most
gas is administered on a one-pass system, i.e. exhaled air, containing most of
the that was in the inhalation, is simply dumped. For most conditions,
pure oxygen is not required, which lowers gas volume requirements to a
greater or lesser extent. Actually, despite the fact that oxygen is funda-
mental to life, pure at pressures of above around 0.5 bar for extended
periods of many hours actually has some toxic effects, in the worst case
leading to blindness.
Oxygen is needed for surgery, for breathing by patients in intensive care
and for breathing by patients with lung and heart disorders where oxygen
uptake is too low. In hospitals, operating theatres and intensive care,
patients are generally supplied by gas from a pipeline system. Oxygen, or
rather oxygen-enriched air, is often supplied in addition in cylinders for
patients needing it in parts of the hospital which are not on the pipeline. For
patients needing high concentrations of oxygen there is no alternative to a
face mask. For those needing only a small enrichment of the air a less
intrusive device is a pair of tubes gently blowing oxygen up the nose.
Continuing oxygen therapy when patients leave hospital is not so easy.
Small cylinders for temporary use during travelling and large compressed
oxygen cylinders for use at home are the norm. For patients who need only a
little oxygen and those who do not need it continuously, this is acceptable.
Oxygen concentrators, small oxygen PSA or membrane units, can be used to
supplant cylinders of oxygen, at least where the concentration needed is
not too high. They are now in regular use for treating people who need
oxygen at home almost continuously, i.e. for more than 16 h per day. For
convenience in the home, and for patients who are away from home for
considerable periods, a liquid oxygen vaporiser system is possible. Already
quite common in the USA, other parts of the world are now looking more
seriously at this. A lightweight Dewar flask can supply oxygen for 8 h at 2 or
31 per minute consumption.
Helium/oxygen mixtures are sometimes used for people with breathing
difficulties. The lower dynamic viscosity of helium means that although
similar in most respects to air, it can be breathed more easily and deeply than
air. The use of helium/oxygen mixtures in diving means that there is a lot of
experience with them. By contrast there is as yet little experience with nitric
oxide, which is now being tried to assist breathing.
Recently, the role in biology, and possibly also in the treatment of
breathing difficulties, of nitric oxide gas has come to the fore. Nitric oxide
can only exist fleetingly inside a biological system because it is highly
reactive. It might at first be surprising that a small, reactive and toxic
molecule should play such a key role. Most chemicals in biology are large,
complex and not especially reactive molecules. It now appears that NO is
used as a chemical messenger for the control of certain muscles, the smooth
muscles which are not under direct voluntary control.
70
NO reacts quickly
with oxygen to form nitrogen dioxide and is then further reacted. However,
it is used as a messenger in biology probably because its small molecular size
allows rapid diffusion and its short lifetime ensures that its action stops
promptly.
Arguably, NO is already important in medicine; the use of trinitro-
glycerine as a remedy for angina pectoris relies on the release of NO. As the
role of NO becomes better understood, its importance in medicine may
increase. Following pioneering work by, among others, Dr Claus Frostel at
the AGA company, NO is now in the process of being trialled in the USA in
the role of relaxing smooth muscles in the lungs to prevent the occurrence of
respiratory distress in newborn babies. The NO is administered via the
ventilator machine, such that controlled low levels of NO are achieved in the
i nfant patients lungs.
N
2
O, cyclopropane and anaesthetic gases. Prior to the advent of anaes-
thetics, surgical intervention amounted to an almost barbaric assault on a
patient and only the simplest operations could be carried out. The discovery
by Humphrey Davy in England of nitrous oxide and the pioneering work of
such clinicians as William Morton in the USA on diethyl ether and James
Simpson in Scotland on chloroform led to an entirely new concept for
surgery, which allowed operations to be performed satisfactorily.
Many gases are too reactive to be breathed at all. However, for those that
are sufficiently inert to be breathed it appears that all act as anaesthetics; it is
simply a question of the pressure at which the gas begins to act as an
anaesthetic and renders the subject unconscious (Table 3.15). People
assisting in a hospital operating theatre are often impressed by the ease with
which anaesthesia is induced. Given the anaesthetic nature of so many
substances this should not, however, be regarded as impressive. The difficult
70
The tale of NO elucidation is an interesting one. The action of NO was first noted by
physiologists working on the erection of the penis.
part, the acid test, of anaesthesia is making sure the patient wakes up again
without after-effects.
The first theorists of anaesthetic action noted that the anaesthetic pressure
of a gas and its solubility in a lipid (olive oil at blood temperature) were close
to proportional. They supposed that the anaesthetics dissolved in the lipid of
nerve cell membranes, inhibiting nerve conduction in some way and thus
inducing anasethesia (absence of pain) and eventually unconsciousness. It
has been shown that the inhibition of nervous signals by anaesthetics comes
about via increasing the permeability of nerve membrane to potassium ions.
The levels of gas used in clinical practice lead to a measurable (n A) ion flow.
The (1961) Pauling hydrate theory ascribed anaesthetic action to the
formation of microcrystals of gas hydrates near nerve membranes but was
not proved satisfactorily. More detailed models of anaesthetic action have
not yet been proved but the current favourite is the FranksLieb model,
which ascribes changes in membrane properties to reactions of gases with
receptor protein molecules in nerve membranes. The theory seems plausible
but is far from being able to predict which compounds will be effective
anaesthetics. There is a large commercial anaesthetic vapour research effort
but this is based on extensive empirical performance and toxicity testing.
A good anaesthetic should be:
potent
non-toxic for the patient (short-term heavy exposure)
non-toxic for medical staff (long-term trace exposure)
substantially non-flammable
easily vaporisable
swift-acting during induction and swiftly removed during recovery
a non-irritant to the respiratory tract.
The latest modern drugs, mainly substituted methylethyl ethers, such as
isoflurane, fulfil most of these criteria well and are replacing halothane,
which in turn replaced chloroform and ether earlier in the century (Table
3.16). However, in hospital practice, these vapours are rarely used alone.

Although practice varies from region to region, the following is a reasonable
guideline:
first the patient is administered an injected anaesthetic, often an opiate,

producing unconsciousness (this is often continued via a drip feed to a
vein);
a further depth of anaesthesia is then induced via a mixture of

and a potent vapour;
if the operation requires it, a muscle relaxant is administered and a

mechanical ventilator applied to the lungs;
unconsciousness is maintained by a low continuous dose of and

vapour, along with oxygen at 20%;
is administered to stimulate breathing during recovery;
the patient is monitored closely during recovery, which occurs as the

injected and inhaled anaesthetics are metabolised or evaporated off in the
breath.
For short operations, such as dental anaesthesia, cyclopropane may be
used alone, whilst for pain relief without unconsciousness, a 50/50 mixture of
nitrous oxide/oxygen at 120 bar (ENTONOX)
71
can be used. The latter is
used especially where, as in childbirth, injected anaesthetics are undesir-
able.
Equipment for administering anaesthetics. In essence a modern anaes-
thetic system still implements the functions of the machines devised by
Boyle and his colleagues in the 1930s. A manifold takes gases and
via regulators to the manifold which mixes the gases, humidifies them,
and delivers them to the patient. A small rebreathing bag may be used
71
It is a curious fact that this mixture exists as a gas in a cylinder at room temperature and 200
bar. If the and behaved independently according to Daltons law, then the should
liquefy whilst the oxygen sits on top. This situation can pertain if an ENTONOX cylinder is not
mixed in the first place or is allowed to stand upright for a period in severely cold (20C)
weather. However, under normal circumstances, an cylinder behaves in most respects
like a cylinder of permanent gas. This is just one example of the breakdown of Daltons law for
gases at higher pressures. Such mixtures, particularly those involving gases which liquefy easily,
are not susceptible to simple theoretical calculations; experimental work is necessary.
before the face mask, whilst check valves ensure that gas flow is unidirectio-
nal. Gas is passed through a vaporiser unit to provide a stream of gas
saturated with vaporised liquid agent at its saturated vapour pressure. The
proportion of each gas supplied is governed by the temperature of the
vaporiser and the setting of the gas flow rates. In simple machines flows are
governed by needle valves and rotameters.
Modern machines provide pin-index gas cylinders to avoid wrong
connections, built-in gas analysis functions (see next section) and provide
separate interchangeable vaporiser modules for different liquid anaesthetic
agents. Inside each vaporiser module is a temperature-sensitive needle
valve
72
which diverts part of the patient gas supply via a tank of vaporisable
agent, which has a wick to ensure that gas leaves saturated with vapour. At
lower temperatures, more patient gas is bubbled through the tank, and vice
versa at higher temperatures. A needle valve on the saturated gas completes
the control circuit, giving the clinician the possibility of varying the set
proportion of vapour agent in gas. An alternative thermostatically heated
tank system, which does not therefore require a temperature-sensitive
needle valve, is used in some vaporiser designs.
The modern anaesthetics trolley also provides facilities for directly
monitoring patient condition with respect to the anaesthetic using measure-
ments of approximate blood oxygen level via a pulse oximeter, breathing
volume monitor and an electrocardiograph (ECG) (Figure 3.47).
Gas analysis in anaesthesia. The closest possible control of anaesthesia is
now pursued in order that the smallest possible dose is administered while
avoiding the patient being even slightly conscious. For all but the simplest of
operations, gas analysers are now in use.
Capnometers monitor exhaled breath, which gives both a breathing
waveform and an idea of the amount of that has been metabolised.
These simple devices are mostly non-dispersive infrared detectors reading
020% More sophisticated analysers, such as multi-channel NDIR
machines, reading and potent anaesthetic agent simultaneously,
are also available. Small electrochemical oxygen cells are inexpensive but
are rarely used because their slow response time (measured in fractions of a
minute) means that they can only measure the average level of exhaled.
The most sophisticated analysers used in the operating theatre are the mass
spectrometer-based instruments. These are typically quadrupole analysers
capable of response times in the tens of milliseconds, better than fast
capnometers, on six or seven gases simultaneously.
Medical mass spectrometers make measurements of peak intensities at a
72
It should be noted that some ordinary industrial needle valves are unintentionally
temperature-sensitive, a feature that can cause problems as set flow rates in apparatus drift off
their intended values with temperature changes.
Figure 3.47 Principle of closed-circuit anaesthesia.
spectrum of different mass numbers, with no attempt at resolution better
than 1 a.m.u. The data is then used in matrix inversion software to
determine gas concentrations. This is necessary because many ion mass/
charge ratios overlap; for example, has the same mass/charge ratio as
Being already in electronic form, the data produced is readily
displayed and, if necessary, processed by the built-in computer to indicate
metabolism and anaesthetic absorption.
On-line blood-gas analysis is also used occasionally in anaesthesia. Simple
two-electrode electrochemical probes inserted into arteries can measure
levels. An alternative means of gas analysis relies on the haemoglobin in the
patients own blood as an indicator. If the patient is receiving too little
oxygen, then the amount of red oxygenated haemoglobin is reduced and the
purple deoxygenated form is increased. Pulse oximeters measure light
transmission (typically at two different red or IR wavelengths) through a
thin part of the patients body and hence oxygenation of the haemoglobin.
For adults the part of the body used is usually a finger or earlobe.
73
Closed circuits. In principle, a closed-circuit anaesthesia system is very
desirable because:
73
For small babies, systems have been devised which look through the top of the babys head,
through the fontanelle or gap in the bone. University College Hospital, London, is one of the
institutes which research this remarkable technique.
smaller amounts of gases escape into the operating theatre;
less anaesthetic gases needed; only that required for make-up of losses
and absorbtion by the patient.
However, such systems have only recently achieved a level of acceptance
and most operations are still performed with once-through gases. The
problems lie with performance of the soda-lime absorber, which scrubs the
produced by the patient, and maintaining the balance of gases in the
system. The latter dictates a higher level of instrumentation (see Figure
3.47).
3.9.2 Measurement of lung function with gas mixtures
Some measurements of lung function are usefully made with simple
apparatus. Huge numbers of people are familiar with the mechanical or
electronic peak flow meter which measures the maximum flow from the
lungs and is used to indicate when inhaled drugs are necessary in the
satisfactory control of asthma. However, with a mass spectrometer or
perhaps a simpler gas analyser, together with a cylinder of mixed gas, it is
possible to make accurate measurements which are more directly related to
lung function.
Oxygen is carried in the blood almost entirely by its reversible reaction
with haemoglobin molecules. So, as it happens, is carbon monoxide, which
forms the stable bright red complex carboxyhaemoglobin, the haemoglobin
being thereby irreversibly poisoned. This latter fact enables a simple test to
establish the oxygen-carrying capacity of a persons (or animals) lung/blood
system. A sample of air with a trace amount of CO is breathed in and the
amount of inhaled CO is measured. Then the reduced amount of CO
breathed out is measured. The reduction measures the capacity of the lung
to absorb CO and hence its capacity to absorb oxygen.
The inclusion in a test breathing mixture of a little helium will allow the
measurement of the volume of the alveoli (the terminal vessels in the lung
from which diffusion into the blood finally takes place) and airways. The He
inhaled is diluted by the ratio of the lung volume to the inhaled volume. By
measuring flow, the latter can be estimated so that total active lung volume
can be derived. This is an important parameter for clinical diagnosis and
treatment. (In many diseases, the lungs fill with fluid, reducing lung
volume.) More ingeniously, diseases where the lungs have trapped air, air
that does not communicate with the airways, can be estimated by
comparison of plethysmographic results with He dilution. A medical
plethysmograph is a large sealed telephone box in which the patient sits and
breathes through a tube. It measures changes in body volume and thus lung
volume (including trapped volume) as a function of small applied pressures.
Boyles law is applied to obtain the total lung volume.
One of the standard lung function tests now carries out both helium
dilution and CO diffusion measurements simultaneously by using a mass
spectrometer and a mixture of 0.3% CO/10% He/21% It
should be added that there are other lung function gases used. Small
concentrations of acetylene or SF
6
, for example, are occasionally used for
similar purposes.
3.9.3 Other medical applications of gases
Cryosurgery. The deliberate use of cryogenic burns for treating some skin
lesions, such as warts, is now widespread. It has the advantage of being a fairly
pain-free process as the liquid nitrogen inducing the burn to the unwanted
lesion also anaesthetises. (Of course this anaesthetic effect wears off as the
skin warms up and cryogenic burns can be as painful as any other.) The
technique is now more or less mature, with specially made applicator heads
and trigger control of the liquid nitrogen flow aiding the medical practitioner
in accurate and effective treatment. The necessity of storing LN in a
superinsulated vessel and regularly topping this up has to some extent
restrained the more widespread use of cryosurgery by general medical
practitioners. The use of alternative freezing fluids has not yet achieved much
success because many are flammable or ozone-unfriendly and, being higher
boiling liquid gases, do not freeze as fast as LN.
An interesting variation of cryosurgery is the possibility of freeze-branding
animals. The application of a liquid nitrogen-cooled branding iron of the
correct heat capacity and conductivity to an animal results in a permanent
alteration in the fur growth, leaving an indelible brand without hurting the
animal.
Gases in surgery. Various minor applications of small amounts of gases have
developed over the years. In some minor abdominal procedures that involve
operating through a small incision, for example, carbon dioxide gas is used to
inflate the abdomen and give a clear view to the surgeon despite the smallness
of the incision. Perfluoropropane and other stable fluoro gases, such as
are injected into the human eye as part of a procedure to deal with detached
retina problems. The gas is injected into the eye during the operation to form a
bubble, which is used to push the retina back into position and prevent fluid
from the back of the retina pushing through any break in it. Once in position,
the detached retina may be welded back in position with a laser beam. Any
mechanical contact with the front face of the retina to push it back into
position would result in damage and must be avoided. The use of a gas bubble
avoids this damage. Perfluoropropane and do not dissolve easily in tissue
and are the favoured gases for this procedure.
Cryostorage. There is little doubt that the cryostorage of humans beloved
of science fiction writers is still a long way away if, indeed, it is ever possible.
However, this does not diminish the undoubted importance of cryogenic
temperature storage of biological samples. At liquid nitrogen temperatures
all normal biological processes are suspended. However, all life depends
on water as its chemical medium and water freezing causes difficulties with
cooling biological samples. Provided the cooling process is carried out
quickly, with, if necessary, anti-freeze additives, many small animals and
plants, and sections of larger ones, can be preserved without appreciable
damage.
Cryogenic dust mite treatment. All around us, but especially in bedding,
live small dust-mites: microscopic insects which feed on minute fragments
such as the tiny pieces of human skin rubbed off while sleeping. The
microscopic faeces of these apparently harmless insects are responsible for
much of the asthma and hay-fever and respiratory tract allergic reactions
seen in medicine today. Most bedding and clothing can be washed
regularly to minimise the amount of allergen to which the asthma subject is
exposed. However, carpets and mattresses are not easily cleaned suf-
ficiently thoroughly. The dust mites cling to cloth with claws which are
powerful relative to their size and even when dead they are not easily
detached.
A treatment which has been successfully applied in Europe for several
years now is to spray the bedding and clothing with liquid nitrogen and,
while the material is still very cold, suck with a high-efficiency vacuum
cleaner. The liquid nitrogen renders the creatures brittle so that their claws
and legs break and allow them to be swept up. Very large reductions in
mite counts are possible with this treatment.
Sterilisation with ethylene oxide. Ethylene oxide is often mixed with
for ease of handling and to render it less flammable. It is a powerful room
temperature sterilisation agent with few side reactions with common
materials and has now supplanted steam in many instances, especially
where the equipment to be sterilised may be damaged by heat. Surprisingly
perhaps, ethylene oxide can sterilise even plastic prepackaged equipment if
the correct packaging is selected, i.e. it has some penetrating power. Pure
ethylene oxide is subject to unforeseen polymerisation reactions under
some circumstances so a shelf-life of 6 months is often specified. Mixtures
with from 10 to 90% are used, as well as mixtures with more ozone
friendly HCFCs such as HCFC-124.
Ethylene oxide has distinct handling dangers, however, and, even in the
small quantities used, is an undesirable addition to our atmosphere. With
the ready availability of oxygen on hospital sites, and the convenience of
computer-controlled ozonisers, perhaps the use of ozone for sterilisation
will become more popular.
3.9.4 Breathing mixtures: diving and aviation
The limits of diving gases. The worlds oceans have many important
natural resources, varying from fish to manganese nodules and petroleum.
However, exploitation of these resources is hampered by the need to access
the ocean floor. Although midget submarines (with their crews either
remote or inside at 1 atm pressure) with manipulator arms can carry out
simple tasks, they are too clumsy for many operations that require a diver to
operate with a thin rubber suit so that he is at the same hydrostatic pressure
as the surrounding water (known as aqualung or scuba operations).
Most scuba-diving takes place in depths of a few metres of water using
compressed air via pipes or, more commonly, in gas cylinders. However,
hydrostatic pressure increases 1 barg for each 10 m of depth and at the
depths usually required for mineral exploitation air is not adequate because
of anaesthesia and the bends. As noted above, pure oxygen is toxic and
pressurised oxygen is even more so. The only solution is to use mixtures of
inert gas with oxygen. The anaesthetic pressures given in Table 3.14 provide
ultimate limits to the pressure at which the diver can operate with these
inert/oxygen mixtures. With helium/oxygen, the deepest diving gas,
anaesthesia can be predicted to occur below 190 bar or about 1900 m. With
other gases diluting the oxygen, lower limits will occur; 240 m for
argon/oxygen, for example.
In practice, other problems occur which prevent aqualung operation long
before these depths. The severe sickness known as the bends occurs when a
diver is decompressed too fast. Bubbles of gas, nitrogen with compressed-air
breathing mixtures, form in the body, especially at joints, and in blood,
causing great pain and tissue damage because the bubbles block capillary
blood vessels. Oxygen and carbon dioxide are metabolised and it is the inert
gases which cause the bends, although helium causes fewer problems than
nitrogen. Viscosity problems occur with higher pressures. At very high
pressures, gas viscosities deviate from the ideal gas (which has constant
viscosity) and begin to edge upwards, increasing flow resistance in the lungs.
Also, the ratio of viscosity/density, the dynamic viscosity shrinks and
formerly laminar flows can change into turbulent flow, also giving rise to
flow difficulties.
74
In practice, diving at depths beyond air depth (100 m or
so) requires helium/oxygen mixtures arranged to give a reasonable partial
pressure of oxygen at the operational depth. A higher proportion of oxygen
is necessary at lower pressures and depths so that the diver is not rendered
anoxic. Rebreathing circuits using soda-lime carbon dioxide absorbers can
be used to economise on helium consumption. Use of helium allows free
74
These flow difficulties are not reflected in many animal experiments. Small animals can in fact
breathe easily, even in viscous gas or liquid.
diving from the surface to 150 m and with acclimatisation 500 m or more is
possible.
Breathing at high altitudes. Higher performance military aircraft and
balloons in the 1930s, and the climbing of very high mountains, quickly
exposed the problems of humans breathing at high altitude. Atmospheric
pressure falls quickly with altitude (Figure 3.48). Although the fall of
temperature with height compensates slightly for the otherwise exponential
pressure fall, it is nevertheless impossible to breathe normally above about
5000 m.
Commercial aircraft today operate in the 12 200m (40 000 ft, subsonic
aircraft) to 18 300 m (60 000 ft, supersonic aircraft) altitude region. At these
altitudes, atmospheric pressure is only 0.19 bara and 0.07 bara, respectively,
too low for breathing. To avoid the necessity for oxygen masks for
passengers during normal flight, therefore, the cabin is pressurised. Other
medical problems which are caused by very low pressure, such as the bends,
are also avoided by cabin pressurisation. However, the pressurisation is not
taken right up to 1 bara but is reduced 25% relative to sea level pressure to
correspond to an altitude of, at most, 2600 m (8500 ft or 0.75 bara pressure).
This slightly reduced air pressure causes little discomfort to passengers and
reduces the stress on the airframe considerably compared to operating at full
atmospheric pressure. Surprisingly, the air pressure reduction actually
reduces the weight of a large aircraft appreciably; by about half a tonne in
the case of a 747 jet.
Oxygen is, of course, necessary for aircraft operating without a pressur-
ised cabin or in the event of loss of pressure in a commercial aircraft. The
latter thus all carry an emergency oxygen system comprising lightweight
cylinders of oxygen and a set of oxygen masks which drop automatically out
of the cabin ceiling. There must be enough oxygen for passengers and crew
to breathe while the aircraft descends normally to 4600 m (15 000 ft, 0.57
bara); the maximum altitude at which it is thought safe to breathe normal air
for a long period.
Manned spacecraft cabin atmospheres have varied with different series of
spacecraft. Russian craft use a mixture similar in composition and pressure
to ordinary air, approximately 1 bar and 20% oxygen, as does the USA
Space Shuttle. As with aircraft, however, it is clearly an advantage to use a
lower pressure atmosphere to reduce stress and the use of oxygen allows
even lower pressures to be used. Early USA spaceflights used pure oxygen at
0.3 bara. After a disastrous fire in an early three-man Apollo mission,
however, mixtures with more nitrogen were used. Collins (1988) records
that the Apollo 1 was particularly hazardous because of the practice of
pressurising the spacecraft to 1 bara pressure whilst on the launchpad. The
pressure was reduced to 0.3 bara only after launch. In 1 bara pure oxygen,
fires are easily started and will consume even normally non-flammable
Figure 3.48 Atmospheric pressure versus altitude.
materials such as steel. The procedure on launch is now to use mixtures with
30 or 40% or more nitrogen at 1 bara, the nitrogen concentration being
reduced after launch as necessary, while more recent spacecraft use
mixtures with more nitrogen anyway (e.g. 28% nitrogen in Skylab and air in
the Space Shuttle).
Acclimatisation is possible for people living permanently at high altitude.
Effects such as the bends are transient and acclimatisation for even a matter
of minutes can eliminate those effects. Other acclimatisation takes much
longer, however. Those living high in the Andes, for example, have very
high haemoglobin counts; their blood has thus an increased capacity for
absorbing oxygen. Athletes in the Mexico and other Olympics held at higher
altitudes have always striven to achieve the maximum level of acclimati-
sation before competing because of this. Even so, record-breaking perform-
ance is slightly less likely at high altitude games. Hunt (1914) describes how a
closed-circuit oxygen-conserving oxygen enrichment apparatus was de-
veloped for use in climbing Mount Everest although the first conquest used
straightforward cylinders. The logistics of getting enough cylinders to a
high enough altitude were difficult. However, the wonderful warm glow of
well-being that a little oxygen gives at these altitudes is well described. The
art and science of acclimatisation have now been carried sufficiently far that
climbs without oxygen are now commonplace; something that would
perhaps have surprised the early climbers. Groups like the Institute for
Aviation Medicine at Farnborough, UK, carry on this research today.
3.9.5 Gases: vital tools of research
Gases of many different types, in the highest practical purity, are a vital
component of many research activities in fields as diverse as nuclear physics
and zoology. Liquid helium, a technical tour de force, is an everyday
commodity in many laboratories today and has made the performance of
almost any experiment that can be performed at room temperature
repeatable at a few degrees Kelvin or so
75
.
Ultrapure gases. Ultrapure grades of gas are important for research. The
problem of using an impure grade is that its impurities must be few,
accurately measured, and of known small effect on a particular experiment.
If they are not, then results may be obtained which are due to an unknown
and uncontrolled impurity effect. It is much the safest course of action in
research to start off with the purest materials, including gases, only scaling
down on these quality requirements when the subject being investigated is
thoroughly mastered.
The use of ultrapure materials in the past has led to discoveries that would
never have been made had research been confined to ordinary grades. The
whole science of semiconductors depends on silicon, germanium and allied
semi-metals being pure at the level of 1 part in 1000 billion.
Fortunately, most gases can be fairly easily prepared in purities that for
most chemicals and materials would be regarded as extraordinary. Impurity
levels in nitrogen at the ppb level are now achieved. However, even in many
other more ordinary purity gases ppm levels of contaminant are the
maximum seen. A common nomenclature for the purity of gases is the
number of nines system. In this system, the number of 9s in the purity
percentage is given, followed by the last figure of the purity. In this way,
ultrapure gases can be more concisely and reliably specified. A gas of purity
4.0 in this system is 99.99% pure, whilst a purity 7.0 gas is 99.99999% pure.
Purity of carrier gases for gas chromatography. High gas purity is necessary
in GC analysis, not least the GC analysis of gases themselves (section 2.3.4).
GC is the standard technique for analysis of vaporisable substances, even
75
The use of even lower temperatures than those of liquid He-4 may even more legitimately be
regarded as a technical tour de force. Although lower temperatures are often achieved by
adiabatic demagnetisation, a convenient refrigerator from 4K down to a few mK is the
dilution fridge. This functions by using the entropy of mixing, as seen in the theoretical air
separation energy calculated in section 2.1.1) to cause cooling. The mixing process used is that
of the exotic (0.00014% of natural He) and expensive isotope He-3 with normal He-4.
those with vapour pressures in the sub-mbar region, and is in many cases the
technique used where some pre-reaction of the substrate can be used to
convert it into vaporisable substances. The latter technique, derivatisation,
allows for GC analysis of involatile substances such as steroid hormones, for
example, in checking athletes for illegal drug-taking.
In using GC for analysing substances of low vapour pressure, however, it
is even more important to ensure carrier gas purity. As a general guide, the
carrier gas should have total impurities less than the desired level of analysis.
To analyse down to the ppm range, a carrier gas in the ppm range or less of
total impurity is probably appropriate, whilst for a cruder analysis at the
fractional percentage level it may be sufficient to use a carrier gas of
approximately 100 ppm total impurities.
Different GC detectors vary widely in their sensitivity to different
contaminants (Table 3.17).
Gas-filled nuclear particle detectors. Many nuclear particle detectors rely
on the detection in special gas mixtures of the ionisation caused by the
passage of a high-energy particle. Some detectors carry out an intrinsic
amplification of the original ionisation. They achieve this by applying a high
electric field, which causes an avalanche multiplication effect. If the
avalanche is limited in extent and duration, and proportional to the original
ionisation, the counter is dubbed proportional. If the avalanche carries on
to produce a local arc then it is known as a spark tube, while if the discharge
engulfs the whole tube, then it is dubbed a Geiger tube. Spark tubes,
operated at atmospheric pressure (often with neon), and Geiger tubes,
operated at lower pressure, are less popular now that electronic ampli-
fication is readily available. A proportional tube with an amplifier recovers
and is ready to detect another particle in a matter of microseconds or less,
whilst spark and Geiger tubes have a high dead time as they have to wait for
charge lost in the large discharge to build up again to operating voltages. The
Geiger tube, however, gives a large electrical pulse and can be made very
simple and reliable, making it suitable for field and educational use.
Most nuclear detection relies on the following principles:
argon as the basic component of the mixture, often at 1 atm pressure

(mainly argon atoms are ionised by the incoming particle);
a hydrocarbon which absorbs ultraviolet light by forming free radicals,

preventing the spread of the discharge all over the detector tube and
giving the detector a fine position resolution (alcohol or isobutane are
common choices);
a halogenated compound such as a CFC, which mops up electrons and

allows faster operation by deactivating the detector more quickly, giving a
shorter dead time;
an optional component (methylal) that prevents polymerisation of the

hydrocarbon radicals to a gunk, which renders the counter inoperable as
it accumulates.
Mixtures of the above basic form, generally containing argon, isobutane,
methylal and CFC for proportional detector tubes, are often called magic
gas. Developments of proportional tube detectors, many pioneered by
George Charpak of CERN, have improved the performance to the point
where proportional detector tubes can count particles at rates of tens of
millions per second using magic gas.
Not all ionisation detectors employ gas in its gaseous form. Effective
detectors for high-energy photons and electrons can be constructed with
liquid argon. Ionisation induced in many liquids can be collected on
electrode plates, although ion and electron transit speeds are lower. Liquid
argon has a reasonably high atomic weight and good electrical and chemical
properties. By adding lead collector plates, a liquid argon detector can be
made very sensitive to the shower of lower energy photons and electrons
produced by an incoming electron or photon in the energy range above
about 1 GeV.
76
Radically different to proportional detectors in principle are Cerenkov
detectors. These rely on the fact that the speed of light is minutely slower in gas
than it is in vacuum. High-energy charged particles travel so fast that they may
exceed the velocity of light in the gas (although still be less than the speed of
light in vacuo). When this happens, a small amount of light is emitted by a kind
of shockwave effect analogous to the supersonic bang of a high-speed
aircraft. The light can be detected in photomultiplier tubes and used to define
velocity limits on transiting particles. The use of dense gases, such as
pressurised CFC, allows the detection of even slow particles, whilst
atmospheric argon only allows the detection of much higher energy particles.
Superconductor magnets and nuclear magnetic resonance. Despite the
burgeoning numbers of NMR or MRI magnets (see section 3.7.5), the
biggest user of superconducting magnets today is probably still the
fundamental particle physics research community. A moving charged
particle moves in a circular path in a magnetic field. At the immense energies
needed to probe fundamental particle interactions, 10 GeV or more, high
magnetic fields are needed to keep the particles in their orbits; even with
accelerator rings several kilometres across. Because these high magnetic
fields must be maintained for hours or days while the particles in the beams
are accelerated and kept in their orbits, superconductor magnets are
indicated. Although there are only a small number of particle accelerators,
such as DESY in Hamburg or Fermilab in Chicago, each contains hundreds
of magnets.
76
Liquid argon detectors have been used in large-scale experiments such the TASSO detector
assembly at DESY Hamburg which discovered the first evidence for the subnuclear gluon
particle in 1979.
4 The future of industrial gases
4.1 Market pull
In order to make useful predictions about the future of any industry, both
sociological and technological factors (market pull and technology push)
need to be accounted for. The recent Technology Foresight study (Calder,
1995) in the UK starts with pulling and pushing factors and works towards a
SWOT
77
analysis of the industrial and research scene in the chemicals sector
and some future predictions. Technology Foresight in the UK and similar
studies in other countries have highlighted some common social and
economic trends such as:

an increased emphasis on technologies with small environmental impact
and big opportunities for technologies which replace those with adverse
effects on the environment;
rapid development of many other state economies towards the highest

industrial level (i.e. the USA) these economies must be expanded with
minimal impact on the environment;
demographic changes towards an older population and smaller families

worldwide, especially in the older advanced economies, giving an
emphasis to healthcare;
products;
the gradual movement of the most advanced economies from large-scale

commodity products towards higher added-value specialised or higher
performance products.
Quite how these general trends will impact on the gases business is,
naturally, unclear. However, some predictions can be made safely. In
services such as power and water, greener as well as more efficient
technologies are indicated. Greatly increased use of electric private
automobiles and other vehicles may provide opportunities for gases in the
production of their batteries. Future electric traction batteries will certainly
be more sophisticated in manufacture and may well be based on completely
77
SWOT: Strengths Weaknesses Opportunities Threats.
the rise of individualism will favour technologies for individualised
THE FUTURE OF INDUSTRIAL GASES 519
novel technology. Increasingly specific pesticides (used in smaller quanti-
ties) may enhance the opportunities for industrial gases in fine chemical
production, as will increased demand for medical drugs.
The growth in the use of high technology methods of food preserving and
the development of effective but fully biodegradable packaging materials
seem certain, given trends towards foods which offer convenience but still
minimise environmental impact and meet demands for taste and minimum
additives content. As more and more women join men at work, and less time
is spent at home, other labour-saving concepts, for example for cleaning,
will become more prominent. It is possible, for example, that processes
based on chemical vapour deposition could be used to deposit dirt-repellent
surfaces (silanisation is already used to some extent) or perhaps an
innocuous supercritical fluid such as carbon dioxide could be used to clean
clothes without the negative environmental impact of dry-cleaning fluids or
soap powders and phosphates.
High-technology materials can enhance efficiency in many industrial
sectors, for example in lighter aircraft structures or higher temperature
engine parts. Similarly, an increased use of biocompatible materials for
prosthetic devices from hip joints to tooth repairs can be expected. Many of
these materials, coatings are a typical example, involve gases in their
production. Coatings and superstrong fibres, despite their apparently dulled
prospects at the moment (lower military spending has had adverse effects),
promise growth, leading to further growth in industrial gases needed in the
future.
4.2 Prospects for new production processes
Across the breadth of production processes the exceedingly low cost of
intelligent computer-based control means that many processes can now be
operated much closer to optimum conditions than has hitherto been the
case. Further decreases in the cost of computing will probably have less
impact. However, the increasing availability of software which combines
sensor readings in an intelligent way and models the process being
controlled at some level will be important in the future. Similarly, software
with a user-friendly graphical interface seems likely to be used on all but the
smallest process plant; here the shrinking cost of computer hardware may
still make a difference since graphical interface software is notorious for its
need for substantial computer power.
There may well be an increased emphasis on sensors in the future; they
have become the weakest link in the chain of optimisation in many cases.
Now that computational corrections are readily available, problems which
have traditionally restricted the use of sensors, such as nonlinearity and
cross-sensitivities, will have less prominence. Computer control will allow
fingerprinting and further widen the use of sensors.
78
Finally, some sensors
which rely intrinsically on computation, such as cross-correlation flow
metering, will become more affordable and more widespread.
If the cost of sensors falls steeply enough, this, combined with the already
low costs of computing and small-bore pipeline components, will allow the
economic construction of smaller gas production systems. Although the
strong scale factors in cryogenic plants will mean there is little impact there,
the impact may be much greater for techniques such as PSA and membrane
separation.
4.2.1 Bulk nitrogen and oxygen
The consequences of the availability of nitrogen and oxygen at significantly
lower prices might well be quite far-reaching. Many processes that are
currently uneconomic with gases would become more economic.
There seems little room for lowering the costs of the liquefied gases by
very much. Liquefaction accounts for the majority of energy costs for LN
and LO, and liquefaction can only be improved by improved refrigerator
plant. This latter point is being addressed in the latest generation of plant
where much higher efficiency liquefiers are now being employed. Beyond
these improvements, and admittedly the modernisation or replacement of
much of todays cryogenic air separation plant will take many years, only
small increments seem possible.
More spectacular improvements may be seen in the lowering of costs for
gaseous oxygen from on-site plants. The separation process is, of course, a
part of the production costs in cryogenic air separation. Cryogenic
separation efficiency has been the subject of a long and sustained drive as
separation plant has grown in scale with the requirements of large gaseous
oxygen and nitrogen applications in the steel and chemical industries.
However, the improvements in efficiency seen in these large plants have not
been matched by the efficiency of smaller cryogenic plants; scaling factors
favour larger plants heavily with respect to separation efficiency. However,
there would seem to be good prospects that the cost of gaseous oxygen
supplied from smaller on-site plant could well decrease, perhaps by 50% or
more given developments in PSA, and it is interesting to speculate what
impact this will have on industrial processes.
78
Fingerprinting is taken to mean the use of sensors to measure and reproduce operating
conditions even where the precise meaning of the measurement is unclear. As an example,
consider the use of a simple quadrupole mass spectrometer (MS). A quadrupole MS will
normally fragment most ionised incoming molecules, with only a few molecular ions, giving a
spectrum with a half a dozen principal peaks. The MS will give a complex mixture of these sets
of peaks, many overlapping, if fed with, say, a hydrocarbon gas mixture from a reactor.
However, the reactor may still be well-controlled by the MS by defining what set of peaks gives
a good result, without knowing which peak corresponds to which molecule.
Humble combustion processes might well be the first beneficiaries,
particularly because the use of pure oxygen can eliminate in exhaust
gases; a major consideration with todays environmental constraints.
Low-temperature, low-pressure processes for converting alkane feedstocks
into more valuable oxygenated products might well become available as
complex biological catalysis chains are brought closer to commercial reality.
Oxygen will provide a useful intensification of these processes, just as it does
in wastewater treatment. As discussed in section 3.3.4, even in high
temperature conversion of petroleum to oxygenated products, oxygen may
have a role to play in avoiding the costs of preheating nitrogen and allowing
simpler recycling conversion schemes.
4.2.2 Pressure swing absorption
PSA/VSA units have now been around for some two decades in more or less
their current form. Their progress into the marketplace is still slow; will it
accelerate as PSA increases in technical performance? Current efficiencies
are still well away from the thermodynamic limit, so there is plenty of room
for improvement. There will undoubtedly be improvements in absorber
selectivity, greater ingenuity in cycle design and some improvements using
vacuum swings. These improvements will all help to lower PSA running
costs. However, it seems unlikely that PSA plants will become much simpler
or lower in capital costs, although some VSA designs show promise here.
Lower running costs may lead to PSA units of the same scale as cryogenic
air-separation units in the future. A pointer to this has been the recent
installation of PSA units of tens of tonnes per day at several small steelworks
(for example, the Irish Steelworks, Ireland). Efficiencies for early oxygen
PSA units has been reported as low as but modern units are
reporting results twice as good at The changeover to
vacuum swing absorption (VSA) in oxygen absorption units has led to
results around
The maintenance requirements of a PSA rig are considerable and the
reliability of efficient compressors in small sizes is imperfect. By compari-
son, a liquid cryogen tank, even with vaporiser, pressure regulation and
valves, is a very simple and reliable system. It could be speculated that if
PSA systems (including compressors) approach the reliability of a vacuum-
insulated tank, even if capital and running costs do not go much lower than
they are now, then they will make much bigger inroads in the marketplace.
4.2.3 Membrane separation
Great expectations have built up around membrane separation. In prin-
ciple, membrane units should provide the perfect solution for gaseous
and at any rate for those applications requiring low to moderate degrees
of purity. The more popular polymer membranes are steadily evolving
better performance, whilst more revolutionary membranes, such as the
ceramic membrane being researched by the Air Products Corporation, may
offer step changes in performance. Current membrane units, however, are
severely limited in performance and are only capable of success in certain
niche applications.
A recent example of the latter is the use of small membrane units for
producing nitrogen in beer cellars. In the dispensing of beer it is convenient
to use nitrogen top pressure. This can avoid rancidity in the beer by
preventing oxygen ingress and can, if used at high pressure, also avoid the
necessity for a pump. Formerly, was used for this purpose. However,
dissolves in beer and can make it too fizzy. can be supplied in
cylinders but this, too, has problems. The cellars below bars, alehouses and
hotels where beer is stored are relatively inaccessible and it is difficult to put
cylinders there. Instead, smaller cylinders are in use that,
although they have a good capacity for in liquid form, contain only a
very limited amount, typically only of nitrogen. Purity requirements
for this application are not high. The high capital and running costs of the
membrane unit are justified by comparison to the relatively high cost of in
small cylinders and the elimination of cylinder handling.
There are other potential applications for membrane separation where
small amounts of air gases are needed. New applications, for which the use
of gas cannot be justified today, may become commercial possibilities with
the simplicity of a membrane separation. Gaseous atmosphere control for
foodstuffs, lightweight oxygen enrichment devices for medical use and
nitrogen for automobile tyres could all use simple, lightweight, membrane
separation.
In general, however, membrane units are likely to continue to have
problems, particularly in larger sizes, with capital and running costs. They
are in principle simpler than PSA but are relatively new and hence of
unknown reliability. To this must be added the reliability problems typical of
running relatively small compressors.
Transit of food under a controlled atmosphere normally requires large
amounts of nitrogen, which cannot be as efficiently stored as (which is
liquid in the cylinders) or oxygen (which is added by adding air). Crude
nitrogen for food atmospheres could easily be supplied by small membrane
units, with all-up weight including compressor perhaps not exceeding that of
sufficient nitrogen storage, although supplying even the modest amount of
power needed to the food storage container will still be inconvenient.
Membranes could also provide the means of removing evolved ethylene to
restrain ripening.
Membranes could be used to generate nitrogen at the point of use in a
factory using the factorys existing air lines. This would have two advan-
tages. First, no dedicated small compressor is needed, instead air from the
existing large compressor on the site is used. Second, no nitrogen pipeline
will be needed; the already installed air lines can be used.
Replaceable membranes might be a way forward in some applications.
Membrane lifetime is always a source of concern in current units. How-
ever, in smaller units, where most of the cost is in other components in the
equipment, a cost-effective approach might be to design in easily replacea-
ble standard modules. Current membrane costs are very high but this re-
flects the fact that membranes are an exotic commodity and made in small
batches. If there was a steady demand for large numbers of identical mod-
ules, as might be the case if a replaceable membrane application caught on,
there would seem to be no reason why costs should not fall steeply because
raw materials and energy for making membranes are not expensive.
4.2.4 Liquid hydrogen
The use of liquid hydrogen (LH) in space vehicles developed the tech-
nology of handling large quantities of LH. Since then, LH has been pro-
posed as a fuel for vehicles. It burns to produce only the most innocuous of
by-products, water, and it is the lightest known fuel for a given energy con-
tent. As discussed in section 3.8.1, a large airliner running on hydrogen
fuel could fly around the world non-stop without refuelling and the BMW
company in Munich, Germany, have demonstrated the possibilities for the
use of LH as an automobile fuel. Safety problems would have to be at-
tacked successfully, however, before usage became widespread. For ex-
ample, perhaps aircraft LH fuel tanks could be designed to dump LH very
rapidly in an emergency, leaving only a small amount of paraffin fuel for
emergency landing. The bulky nature of LH (it is only one tenth as dense
as water), however, as well as its flammability may prevent its use in
aviation.
Predictions have frequently been made in the past about the rosy future
of fuel cells. It is certainly true that hydrogen-fuelled cells are now reason-
ably reliable and efficient (although it is worth noting that fuel cells were
one of the most troublesome pieces of equipment in earlier spaceflight mis-
sions such as the Apollo expeditions to the moon). Perhaps advances in the
technology of fuel cells for ordinary fuels such as methane, propane or
even petrol will come about. Currently, however, cells using hydrocarbon
fuels are less efficient and reliable, and even more expensive to make, than
hydrogen fuel cells.
With modest improvements in lowering costs and improvements in life-
time and reliability, fuel cells would make adequate power sources for
transportation leaving the hydrogen supply as the difficult factor. The
Hindenburg factor still haunts even though handled correctly would
probably be no more dangerous than petrol. Also, hydrogen liquefaction
will always be expensive, for simple thermodynamic reasons. Hydride
storage is an alternative to LH which may have a role to play.
4.2.5 More efficient gas cylinders
The first non-military uses of wrapped gas cylinders are now starting up.
Whether these small beginnings presage a major move towards wrapped
cylinders is a moot point. The wrappings, although apparently robust
enough, may prove to be too easily damaged by sharp point impacts, either
on customer premises or during filling and transit (the gas cylinder must be
one of the worlds most abused objects).
79
The very slow trend towards the use of higher pressure in cylinders seems
likely to continue. Although higher pressure cylinders are not necessarily
more efficient in gas/packaging ratio over larger but thinner-walled cylinders
at the same pressure, the convenience of a smaller size, and a more robust
thick-walled container, means that the higher pressure cylinder will
continue to be favoured. As more standard gas equipment becomes
available at 300 and 400 bar, the dodge of adding a cylinder valve
preregulator may be avoided and the newer higher pressures may become
standard. The whole process, however, is restrained by the availability of
stronger steels and alloys and by the largest size and weight of object that can
be conveniently handled. A cylinder should not exceed 100 kg or so in
weight and 1.5 or 2 m in length.
At the other end of the scale from ultrahigh pressure are cylinders storing
gas with absorbers present at low pressures. This is a technique which seems
ripe for use in many gases. The applications to dangerous semiconductor
hydrides and hydrogen may point the way. It would be possible, for
example, to use zeolite absorbers on special calibration gases and mixtures
used in instruments (such as mass spectrometers) which can easily remove
gases from a cylinder at subatmospheric pressure. With subatmospheric
pressure storage, in addition to safety advantages, valuable gases such as
xenon would not be lost to inadvertent leaks; an economic advantage to the
user.
High-pressure absorber systems, other than the storage of acetylene in
acetone, may also have their place in the future. For example, vehicles
burning methane as fuel have both economic and environmental advan-
tages. Methane, however, must be stored in heavy, low-capacity high-
pressure cylinders whereas propane, for example, can be stored in low-cost,
lightweight cylinders in liquid form. These disadvantages could be offset by
ultrahigh pressure cylinders. A more economical solution, however, might
79
I have seen gas cylinders variously used, for example, as rollers for moving heavy objects and
traffic bollards!
be found in some sort of solid or liquid adsorber for methane, for example an
enhanced hydrate compound.
4.2.6 Carbon dioxide
As more and more factories and power stations are forced to clean up their
exhaust stack emissions, will be easier to extract from stacks. The new
cleaner stacks will remove substantially all the sulphur dioxide and nitrogen
oxides that currently make extraction of from stacks expensive. It is
therefore possible that may stage a gradual come-back in Europe and
challenge at the lower quality end of the freezing/chilling sector.
4.2.7 Acetylene
The slow decline in acetylene sales has made R&D of more efficient means
of production and delivery for the gas an unattractive area for industrial gas
suppliers. The use of acetylene by many small users for welding, however, is
not contracting as fast as many had predicted. In addition, acetylene will
continue to be used for many years by large numbers of users, both large-
and small-scale, for oxy/acetylene cutting. The industry could well rational-
ise further on fewer primary sources of the gas for commercial reasons.
Some companies will choose to stay with acetylene production and others
will buy it in. If an effective new production process were devised which
showed economies of scale stronger than the current processes, this would
be especially desirable and profitable for the companies remaining in
production. Similarly, improved means for delivering acetylene (smaller,
lighter cylinders with increased contents) might succeed in enhancing the
attractiveness of acetylene to users and holding up sales of acetylene.
4.3 Developing new applications
Developing new applications for gases is an old-established practice in the
industrial gas industry. Ever since the oxygen pioneers such as the Brin
brothers ran out of limelights to supply, new applications for gases have
been sought. New production processes enable the production of gas in
larger quantities at lower cost. Supply then outstrips demand and new
applications need to be sought. All the gas majors have gas application
laboratories, often duplicating the facilities in several major countries where
they operate. The application laboratories often fall somewhere between a
true R&D facility and a technical sales and sales support operation. This
combination of supplier push along with customer pull has proved a very
fruitful way of advancing technology in the past and it is a feature of the
industry that will most likely be retained even with major business
organisation changes. Fortunately, the structure of the industrial gas supply
industry is such that most suppliers are large businesses that should be able
to afford to run adequate development facilities for new applications.
It is always dangerous to try to assess what the future will bring but any
prediction of new gas applications is yet more fraught because of the vast
field of industrial enterprise it must encompass. What follows is a selection
of applications that may be important in the future.
4.3.1 Oilfield injection
Although the principles are now well known (section 3.3.3), oilfield
injection is currently only employed at a very few sites and then often only on
a trial basis. It has the potential of being an enormous user of gas, probably
nitrogen, if it becomes a more widespread practice. As enhanced oil
recovery itself becomes more widespread, gas injection should become
more common.
The availability of very compact, very high-capacity nitrogen generators,
however, may well become a factor in oilfield injection. With an increasing
number of oilwells operating on platforms offshore, the size of plant is a big
issue because of the high cost of space on the platforms. Even with oilwells
on land, injection may only be required for a period of months or a few
years, not the ten years or more required to achieve payback on a large
cryogenic air-distillation unit. Smaller separation plant is thus indicated on
portability grounds.
With the cost and size of PSA plants due to decrease, even for very large
sizes, and with purity requirements of the application minimal, it seems very
likely that they will be used more widely. If membrane separation units
prove possible to scale-up drastically, perhaps they will provide a solution.
The huge size of large-scale cryogenic separation units would appear to rule
them out but perhaps concepts such as the centrifuge distillation tower could
so shrink the size of a distillation unit that cryogenic methods, already
proven at the 10 000 tonne per day level, could be adapted to offshore oil
recovery use.
4.3.2 Transonic wind tunnels
There are now two large transonic wind tunnels, the ETWT (European
Transonic Wind) tunnel, Cologne, Germany, and the NASA facility at
Langley, Virginia, USA. These facilities avoid the need for still larger
facilities. By boiling-off large tanks of liquid nitrogen and operating at low
temperatures (90 K), they can operate at lower viscosity and higher density
than an air tunnel, allowing higher Reynolds numbers while keeping the
tunnel a reasonable size. The linear shrink factor enabled is around or
about 6, which gives a volumetric reduction of 200 on the size of the facility
needed. Where it is necessary to reach Re values of several million, needed
for looking at the complex transonic flows around a large airliner, for
example, this kind of shrinkage is vital. More wind tunnels, perhaps on a
smaller scale, can be expected in the future although the market for such
complex research tools is limited.
4.3.3 New applications for liquid nitrogen freezing
Shortly after the invention of the laser, it was described as a solution looking
for a problem. Today, lasers are being used in many processes but new
applications are still being found almost daily. Similarly, the low cost and
almost ubiquitous availability of liquid nitrogen (LN) is a technology
looking for a new problem to solve. New applications of LN are almost
routine; there is a constant stream of new inventions. There would seem to
be no reason why this should not continue in the future, too. New processes
will always be found which require the ultrafast freezing, cleanness and
convenience that LN freezing conveys.
4.3.4 Supercritical carbon dioxide
The possibility of supplanting organic or organochlorine solvents with
supercritical was mooted in section (3.4.8). Further to this, there are
many industrial reactions carried out in organic solvents which could
perhaps be run in the future in The cost of is lower than organic
solvents, so solvent recovery might not be necessary, simplifying processing.
Hydrogenation and polymerisation reactions are both candidates for
supercritical processing.
The way in which food is processed today is increasingly being driven by
consumer pressure to achieve higher quality, lower cost and lower
environmental impact. Could supercritical fluid extraction (SFE) be em-
ployed in the extraction of oils and flavourings from crushed plant material?
SFE could offer higher efficiencies and purer product than simple crushing
or heating but without the use of organic solvents.
In a similar vein, perhaps, one stage of wood pulp processing could be
carried out using SFE, extracting undesirable wood components such
as polyphenolics. The presence of these compounds gives paper made from
such wood pulps a dark colour. Pretreatment with methanol or other
solvents can improve brightness but SFE might offer advantages in cost
and environmental control. Another application of supercritical with
wood has recently been reported: the preservation of ancient waterlogged
wood (Kaye 1995). The steadily increasing importance attached to the
preservation of historic and prehistoric artefacts means that further
inventions which employ industrial gases in museum conservation work can
be anticipated.
4.3.5 New developments in steel production
Even small changes in the way gases are used in steel production may have a
large impact on the gas industry because of the large size of the steel
industry. Entirely new processes in iron and steel may yet evolve. Direct
reduction, for example, was unheard of until the last decade. However, in
the main, technical possibilities that already exist will most likely be further
developed, driven by the environment in which the manufacturers find
themselves. In addition to the usual drives for greater energy efficiency, a
number of factors may drive the steel industry to change its methods
significantly over the next decade:
1. the slow change of large-scale steel production from (in most countries) a
government-run monopoly, or at least a subsidised commercial oper-
ation, to a truly commercial business;
2. the modern emphasis on low pollution;
3. the modern emphasis on high levels of recycling;
4. the slow growth in importance of the economies of developing countries.
The last point means that larger and larger volumes of scrap steel will
become available, reducing demand for pig iron from blast furnaces and
perhaps restraining the growth of direct reduction iron, at least for larger
plants. Scrap steel allows for an increased growth of mini-mill operations
and mitigates against the leviathan integrated blast furnace/steel refining/
rolling mill complexes that have dominated steel in recent decades.
Change away from subsidy may also encourage mini-mill operations by
encouraging private capital, often of more modest proportions, into new
steel plants. However, breaking away from parochial nationalism may lead
to greater rationalisation in large-scale steel production and may mean that a
few even larger integrated plants could be constructed in the future to
replace the current larger number of smaller integrated units.
The increasing use of scrap will increase the problems of tramp elements
(unwanted impurities) and heavy metals. Perhaps novel innovatory ap-
proaches using gases will be devised to help deal with these.
An insistence on low pollution in new plants will put strong pressure on
new processes to avoid the use of coke. Coke-making plants are now
considerably older than their design life in many parts of the world and when
these plants are closed they are likely not to be replaced. New iron-smelting
methods will instead avoid the need for coke in new plants. The removal of
coke from blast furnaces is also being driven by processing cost and
shortages of suitable coking coal.
Direct reduction of iron ore to iron is one method of removing the need
for coke. An increased number of mini-mills, even if these are largely fed
with scrap, may increase demand for direct-reduction feedstock to smooth
fluctuations in scrap quantity and quality. Direct reduction, although
apparently simple, is still not a mature technology and some further
innovations can be expected. A new process is being claimed by the
NUCOR/US Steel joint venture in Trinidad based on the reduction of iron
ore to iron carbide rather than to spongy iron, the usual product of direct
reduction (see section 3.1.3). A mixture of fines, finely powdered ore, in
this case from Brazil, is heated to 600C in hydrogen and methane. The
result of this is a powdered iron carbide containing 6% C approxi-
mately), which can be heated to drive off some carbon, then melted and
refined into steel using the heat generated by the lancing oxygen/carbon
reaction in the usual way.
The implications of all these changes for the use of gases in the industry
are clear in some cases, less so in others. It is clear, for example, that oxygen
use in blast furnaces will increase as coke declines. The increasing numbers
of mini-mills will lead to more demand for argon for plasma-melting and for
ladle stirring.
4.3.6 Changes in the liquid fuels market
Small changes in the liquid fuels market could also have a large impact on the
demand for oxygen and other gases. The South African SASOL project,
with its large oxygen plants, shows how big the demand for changes could
be.
The use of oxygen in the manufacture of methanol from syngas is
currently not a particularly important route to liquid fuel. Methanol is barely
accepted as a fuel owing to its current high cost and poor calorific value.
(Although, paradoxically, methanol can be used to extract higher powers
from internal combustion engines.) However, as oil prices rise, coal, still
currently decreasing in production, may once again become more important
and its conversion to syngas and thence to methanol offers a relatively
elegant route to a convenient, clean-burning liquid fuel. The scale of such a
change, if it ever happens, would be enormous. The use of oxygen would be
dictated by the reduction in methanol plant size that oxygen can give,
otherwise investment needed in the new fuels would be simply too high.
4.3.7 Nitrogen tunnels for semiconductors
There is currently a movement in semiconductor manufacturing plants
(wafer fabs) away from the use of large, extremely expensive, Class 1 or
better cleanrooms and towards the mini-environment concept. This
concept involves using an inexpensive Class 1000 standard of cleanroom but
with the silicon wafers being processed inside machines, transfer tunnels or
transfer boxes (the mini-environment) where the atmosphere is controlled
to better than Class 1 conditions. The mini-environment approach has been
taken by several recent wafer fabs, for example GEC Semiconductors in
their Roborough, UK, facility and by TSMC in Taiwan. These new facilities
are promising very good conditions for high yielding silicon integrated
circuit (IC) processing but without excessive cost and with a more pleasant
environment for the operators manning the cleanroom. (Class 1 or better
cleanrooms require special uncomfortable clothing and special helmets if
contamination of silicon wafers by the operators is to be avoided.)
Beyond mini-environments a further development can be envisaged. In a
truly ideal semiconductor cleanroom, silicon wafers would not see air, even
highly purified air, but only pure or vacuum. They would travel from
machine to machine in a nitrogen tunnel. The mini-environment of each
machine would be nitrogen or possibly vacuum. There is still debate on the
relative desirability of and vacuum. The leading ultraclean group of
Professor Tadahiro Ohmi at Tohoku University, Japan, prefers nitrogen.
Some researchers in the industry think that vacuum is inferior in perform-
ance but often superior in practice because it is better matched to the mostly
low-pressure processing chambers of a typical wafer fab. The argument for
revolves around it providing protection against the almost instantaneous
diffusion of contaminants which occurs in vacuo. Perhaps for a whole
wafer fab process may not be necessary but islands or clusters of handling
will be used, together with other clusters where handling is in vacuum.
There are hints that some new research installations will use high purity
backfilled into a low vacuum, e.g. 1 mbar. This is thought to give an
optimum low particulate and gaseous contamination. Mean free paths for
contaminant vapour molecules are still short but particulates will fall quickly
and be extracted by the continuous flow vacuum pumping/gas supply
system. It is certainly true that whatever the pressure, flowing is ideal.
Today it is not always the case that the atmosphere is the major source of
wafer contamination. The processing machines themselves, some gas
processes, many wet processes and many photolithography processes all
yield contaminants, and as the wafers shift to smaller and smaller feature
sizes, smaller and smaller contaminant particles become important. How-
ever, more attention is being paid to making the processing equipment
cleaner, which will therefore focus more attention on achieving a cleaner
wafer environment.
Wafer tunnel handling is by no means the only way to achieve the required
goal of an ultraclean mini-environment for silicon wafers. Integration of
many wafer processes onto one machine or suite of machines linked by a
vacuum handling robot arm is an obvious alternative. Creation of a cluster
tool is useful for local handling. The use of ultraclean versions of the
so-called SMIF boxes or similar is another route. Here the principle is to
employ machines that remove wafers from a sealed box, process them and
return them to the sealed box. The SMIF box is then moved from machine to
machine, not the wafers themselves. Further sophistications involve replac-
ing the passive SMIF box with an active module. This blankets the contained
wafers in pure air or nitrogen, which is recirculated via a filter to minimise
pick-up of contamination from the SMIF box itself. A project called
WAFEC, part of the pan-European JESSI project, is studying this at the
LETI microelectronics institute in Grenoble, France. Vacuum is again an
alternative, and the MMSTgroup at Texas Instruments, Dallas, Texas, uses
vacuum SMIF-type boxes.
4.3.8 HTc superconductor cooling
A large range of devices are projected to be made using the new ceramic
oxide-based superconductors, such as 123 YBCO or 2223 BISSCO.
These materials have the remarkable property that they are supercon-
ductors at relatively high temperatures: up to 140 K has been claimed. They
also are capable, when cooled to liquid helium temperature (4.2 K), of
resisting much higher magnetic fields than superconductors.
The first superconductor devices just appearing in the market place
are mostly liquid helium-cooled.

High current leads: connect superconducting coils to room tempera-
ture power supplies. The lead conducts current but not heat and any
heat influx to a helium-cooled superconductor is very expensive.
High-field magnet inserts: coils placed inside coils, operated at

liquid helium temperatures. In this way, magnetic fields as high as 30 tesla
can be reached. These high fields are especially interesting to NMR
chemists (section 3.7.5).
These are very specialised applications and will lead to little additional use
of liquid helium or liquid nitrogen. However, a second generation of
devices using liquid nitrogen cooling seems likely in the foreseeable future,
perhaps by the turn of the century.
Large powerful magnets: for applications from nuclear magnetic

resonance imaging (hospital body scanners) to mineral extraction.
motors and generators: rotating electrical machines all benefit from

superconductors. Many different superconductor machines have been
made with liquid cooled material, some on a very large scale, but
they have so far all had problems which have made them uneconomic.
Fault current limiters: for use as high-speed fuses on high-voltage

electric power lines. Varying the gas pressure above a bath of liquid
nitrogen containing a piece of superconductor acting as a fault
current limiter is a possible way of varying the critical current of the
superconductor and hence the current limit.
80
Once thought impossible, technically very difficult LHe-cooled cables
have actually been constructed and run for specialist purposes. LN cables
will therefore certainly be practicable if superconductors become
practicable large-scale conductors. The economics of LN-cooled cables are
now well-known. Basically, all high-powered cables that have to run
underground would be replaceable if HTc materials go low enough in cost.
Sections of cable, once initially cooled, would be maintained at 77 K by small
ASU/refrigerator/cryogen pumping units located at intervals along the
underground run of the cable.
4.3.9 Diamond production
The production of diamond thin-film out of a mixture of gases is a perfectly
practical, although as yet not fully developed, technology.
The know-how for what is now the standard process for the production of
artificial diamonds was developed over 30 years ago by General Electric in
the USA. It involves the production of stupendous pressures and high
temperatures in a tiny reactor volume. Although now a well-established
technology, it can produce a product adequate for specialised uses but at a
cost higher than that of natural diamonds. The possibility of coating tough
substrates with diamond films was also realised at this time but remained
intractable until techniques and equipment borrowed from the semicon-
ductor industry improved the control of diamond formation to the extent
where consistent product could be produced. As know-how has grown,
deposition rates have gone up, from approximately one micron per hour
with early CVD up to a maximum of one hundred microns per hour today.
In the thermal method (e.g. Hirose, 1987) a hot tungsten filament
decomposes the gas mixture at low pressure (tens or hundreds of millibars)
and a heated substrate (hundreds of C) is placed below this. All sorts of
gases can be used, including many alcohol and ether vapours, methane and
acetylene. Plasma-assisted diamond thin-film production relies on the
production of methyl radicals and hydrogen atoms in a microwave plasma
discharge. The substrate is heated in the low pressure chamber to 800C or
more. The method using an ECR (electron cyclotron resonance) plasma is
similar but uses a magnet as well as microwaves, giving a more intense
plasma. Many vacuum reactor makers now market systems specifically for
diamond film deposition, which can deposit several tens of microns per hour
of diamond, whilst a smaller number of suppliers offer plasma-jet systems,
80
In Downie (1991) HTc material was coated in an enhanced boiling surface to ensure that it and
cryogen temperature would track very accurately the temperature corresponding to the applied
pressure.
which can deposit several hundred microns per hour. A plasma-jet
deposition system might run just below atmospheric pressure with argon/
hydrogen/methane.
Purity issues are important in gases used in semiconductor CVD, with
levels above a few ppm provably affecting film quality. In an analogous way,
purity may well become an issue in achieving higher quality diamond films.
Even isotopic purity is important with diamond films. Films made with
isotopically pure methane have much higher thermal conductivity than
those made with normal methane containing 1% actually sell
methane products for this reason.
Practical, though expensive, industrialised technology now has a large
turnover, with products enhanced with diamond films worth billions of
dollars per year. As the new equipment allows the more widespread use of
the technology, this business looks set to grow much further. Other hard
coatings are used industrially, such as SiAlON, BN and SiC, and these are
being researched and used to some extent in the engineering industry.
Research groups are currently looking at CVD or PECVD (plasma
enhanced CVD) techniques for laying down very high quality films of these
materials and growth can be expected for these coatings, too.
Applications for hard coatings are mostly obvious, e.g. drill bits, wear
resistant parts and bearings. One of the principal applications for diamond is
less obvious: heat-sinking microelectronic circuits. This relies on the high
thermal conductivity of diamond, which is much higher than that of
the standard microcircuit thin film insulator. Also, doped diamond is a
semiconductor and it is conceivable that thin-film diamond semiconductor
devices will one day be produced commercially.
4.3.10 Production of superstrong whisker materials
It ought to be surprising to anyone who has studied chemistry how weak
most materials are. After all, if the tensile strength of a single chemical bond
is of the order of then materials should have tensile strengths of the
order of 100 GPa
81
, whereas typical strong materials, such as steel, are only
0.5 GPa in standard form, although 2 GPa or so can be reached in special
hard-drawn alloy wires.
When a material is stretched cracks tend to form, concentrating stress in a
small part of the material, so that it fails long before the average stress is up
in the GPa region. One way of controlling this crack propagation problem
and producing much stronger, lighter materials is to fabricate composites. In
these strong fibres or whiskers are held together in a matrix, such as a
polymer, of modest strength. Cracks cannot propagate from one fibre to
81
For example, at 100 GPa an ordinary automobile weighing 1000 kg could be suspended by a
thread only 0.2 mm in diameter.
another and the matrix is insufficiently stressed to fail. Tensile strengths can,
therefore, be high.
Two types of high-strength fibres are now in routine production: those
based on aramid plastics, such as Kevlar and those based on carbon fibre,
such as Graphil. These are now of sufficiently low cost that they are used
for high-value consumer items as well as in industrial and aerospace
applications. Tensile strengths of up to 4 GPa for aramid and carbon fibre
materials are possible. These are high strengths for such lightweight
materials as opposed to the of steel). Even
higher strength is possible with short fibres or whiskers of other materials
such as boron or silicon. In addition, these fibres, being stable at high
temperatures, can be used as the reinforcement of a metal matrix.
Carbon fibre is often produced from polyacrylonitrile plastic fibres by
stretching, heating in hot air, then pyrolysing in inert gas. Boron and silicon
carbide and superstrong whiskers are made by CVD deposition from
diborane and silane mixtures, respectively, onto fine tungsten wire,
wires are built up to or so in typical processes. The production of
ultrastrong materials is dependent on defence demands. So far only military
hardware needing the utmost strength/weight ratio has justified using these
expensive materials. They are currently made in very small (and decreasing)
amounts and suffer from the high costs of all small production rate materials.
However, breakthroughs in ultrastrong fibre production, similar to those
seen in carbon fibre, may well be possible and this would lead to a major
expansion of gases used in this sector.
4.3.11 Greener processes
With the growing realisation of the importance of environmental issues, it
seems highly likely that many new applications for industrial gases will stem
from greener processes in industry. A myriad of small industrial processes
will switch to greener technology, often involving gases. One example is the
use of solvent-free coatings (section 3.8.13). The use of solvent-free coatings
relying on radiation curing is expected to grow in the future and will have the
benefit of reducing the amount of solvents emitted to the atmosphere
causing local and regional pollution. With lower cost nitrogen from
membrane and PSA units conveniently available, the use of superior
coatings (which require inerting during UV curing) will be possible on a
larger scale.
Similarly, changes expected in major industries will lead to increased
opportunities where industrial gases can contribute in processes. Some of
the processes already described above will achieve major environmental
benefits (Table 4.1).
Biochemical methods also offer a route to greener processes where they
can supplant methods currently using bulk chemicals. In the case of chemical
synthesis biochemical methods can offer good yields and specific isomers
where classical methods yield mixtures of isomers, some of which may be
pollutants. In other industrial areas, biochemical processes can offer a
greener solution by avoiding other pollutants.
In many biochemical processes some industrial gases may be needed. The
reactor or fermenter vessel will need to have pressures of gases, such as
oxygen, carbon dioxide and inert gases, carefully governed by removal or
addition as appropriate. In biochemical methods which rely on solid plant or
animal tissue cultures or thick beds of single cells, diffusion oxygenation of
the cells may be easier with pure oxygen rather than air. In many such
processes it will be an advantage to use pure oxygen to accelerate reactions
and enable smaller fermenter vessels to be used. One example where
intensification with oxygen might be needed for an economic process is
oxygenated hydrocarbon production, which could conceivably convert to
biocatalytic methods in the future using enzyme-derived catalysts at room
temperature. The size of fermenters may become an issue with the more
widespread adoption of biochemical methods. For example, the ICI
single-cell protein unit at Billingham, UK, is over 100 m high and 10 m in
diameter.
4.4 Epilogue
Industrial gas production technology is now so well-established that there
would seem to be little room for major innovations. The unspectacular but
steady growth of alternative production processes such as PSA in recent
years perhaps points the way. Large changes may eventually be achieved but
these changes will come over a considerable period of time.
Industrial gas technology is an important and still growing part of the sum
total of industrial technology. Whilst the changes in industrial gas consump-
tion figures expressed in tonnes of gas used may still be dominated by
relatively small changes in the big consumers in the iron and steel industry,
there are many other areas of industry which will see significant changes
towards processes using industrial gases. These process changes will affect
the gases industry considerably but will have an even more far-reaching
impact on the users of industrial gases.
The story of gas technology has had many surprising twists in it and it is
hoped that the reader will have gained some appreciation of these by reading
this book. Gases are now a maturing industry in many ways. It is certain,
however, that the technology of industrial gases still holds some surprises for
the future.
Appendix A: Glossary
Gas industry terms
Absolute pressure Pressure referred to zero pressure or vacuum, as
opposed to gauge pressure.
AOD Argonoxygen decarburisation: a process involving blowing argon
and oxygen through a ladle of molten iron. It is applied to reduce the carbon
content of steel.
APU Argon purifier unit: typically used to describe units which heat argon
to a few hundred C and pass it over a supported copper reducing agent,
where oxygen contamination is eliminated.
ASU Air separation unit: normally a cryogenic plant for separating the
principal constituents of air, i.e. nitrogen, oxygen and argon.
Atmosphere In the gas industry atmosphere can denote any gas-filled
space, such as the inside of a furnace, as well as the ambient air atmosphere.
Bar, bara, barg The principal modern unit of pressure measurement in
industry. The earths atmosphere exerts an average pressure of 1.013 25 bar
at sea level. Bar(a) or Bara means pressure referred to zero pressure
(vacuum). Bar(g) or Barg means pressure referred to (local) atmospheric
pressure. 1 bar = 14.5 psi.
Blanket A gas blanket typically refers to a volume of inert gas that shields a
sensitive part of a process from damage caused by air.
Bursting disc A safety relief valve consisting of a thin stainless steel
diaphragm, obtainable in various thicknesses that rupture at different
specified pressures.
Carbide Calcium carbide is reacted with water to make acetylene.
CFC Chlorofluorocarbon.
Coal gas The gas produced by the pyrolysis of coal.
Compressed gas A generally recognised legal term for any compound
which exerts a pressure of more than 0.5 barg on its container at room
temperature.
Cryogen Any liquid boiling below c. 110 K.
Decant A technique for delivering gases compressed in bulk in which a set
of cylinders on the delivery vehicle are allowed to equilibriate in pressure
with the users bank of cylinders.
Dewar A small (less than 2001) vacuum-insulated vessel for storage of
cryogenic liquids at low pressure.
Dip-tube A tube, typically of plastic, copper or mild steel, running from
the valve at the top of a gas cylinder to the bottom. A dip-tube can help with
the mixing of compressed mixtures during the filling process. It also enables
the user to withdraw liquid from the bottom of a gas liquefied by the cylinder
pressure without turning the cylinder upside down.
Doping gas A gas, typically a hydride or organic derivative of an atom from
Group III or Group V of the Periodic Table, which is used to add controlled
amounts of those atoms to a semiconductor to render it suitably electrically
conductive.
Filling ratio Used in fixing the maximum fill of gas cylinders, this is the
number of kilograms of gas added divided by the cylinder volume in litres.
Flashback flame arrestor A device to prevent propagation of a flame
backwards down a gas pipeline from the burner and towards the gas source.
Gauge pressure Pressure referred to ideal atmospheric pressure, 101 325
pascals, or about 15 psia.
Goose-neck eductor A small dip-tube enabling liquid or gas withdrawal,
depending on orientation, from a compressed gas cylinder that is held in the
horizontal attitude.
GWP Global warming potential: the capacity, relative to of a gas for
causing global warming via the infrared greenhouse effect.
HCFC Hydrochlorofluorocarbon.
HFC Hydrofluorocarbon.
Kieselguhr The porous silica material (the fossilised remains of micro-
scopic diatoms) used to absorb acetone in acetylene cylinders.
LC50 The concentration by volume of gas in air that will cause the death
within 14 days of 50% of a batch of male and female 200300 g rats after 1 h
continuous inhalation.
LN/LIN Liquid nitrogen.
LO/LOX Liquid oxygen.
LPG Liquefied petroleum gas: covers both commercial propane and
butane and mixtures of these.
MAP Modified atmosphere packaging: replacement of air in food packag-
ing with a more suitable mixture of harmless gases.
MAPP Methyl acetylene propadiene: a vaporisable liquid used to augment
flame temperature and flame emissivity of the lower performing oxy-fuel
mixtures.
MCP Manifolded cylinder package: typically 15 standard (c. 7
compressed gas cylinders mounted in a frame with a manifold giving one
outlet pipe weighing approximately 1 tonne.
Mercaptans A range of analogues of alcohols in which the OH group is
replaced by the SH group. They have an exceedingly strong smell and are
typically added to heating gas in ppm quantities to enable people to smell gas
leaks.
MIG welding Metal inert gas welding, electric arc welding in which a
APPENDIX A: GLOSSARY 539
consumable electrode melts and is incorporated in the welded joint. The inert
gas prevents the incorporation of too much oxygen in the joint.
MMA Manual metal arc-welding with stick electrodes without the benefit of
gas shielding.
The industry standard gas unit (except US), a cubic metre of gas at
normal temperature and pressure.
ODP Ozone depletion potential: the amount by which a gas will cause ozone
depletion in the stratosphere relative to freon-12.
OFN Oxygen-free nitrogen: a reasonably pure grade of nitrogen that has
been filled into a preevacuated cylinder. In principle a cylinder of nitrogen at
200 barg, filled on top of an empty cylinder, actually containing 1 bar air,
would contain 1000 vpm oxygen. In fact ordinary nitrogen cylinders are rarely
as impure as this.
Permanent gases Gases which cannot be liquefied by application of pressure
at room temperature. H
2
, He, N
2
, O
2
and Ar are the common permanent
gases.
Plug flow Occurs where gas A flows into a pipeline to displace gas B with a
well-defined front, like a piston displacing gas.
ppm/vpm/ppb ppm = parts per million, usually by weight, vpm = parts per
million by volume. The term ppm is often used carelessly, when vpm is in fact
meant. vpm is more useful in the sense that for gases the volume ratio is
approximately a ratio of the number of molecules and is a partial pressure
ratio. ppb = parts per billion, i.e. parts per 1000 million.
PPU Prepurifier unit: typically used in connection with purifiers for air in
cryogenic ASUs.
Producer gas A flammable mixture of and produced by the
action of steam and air on coke or coal. It is common on large integrated
chemical plant. The air ensures that the reaction is exothermic and can be
more or less continuous. However, the quality of the gas is very poor with
often over 60% consisting of nitrogen.
PSA Pressure swing absorbtion (see section 2.1.6).
psi, psia, psig Pounds per square inch, a unit of pressure still in common use
in industry. psi(a) or psia is referred to vacuum. The most common unit is psig,
which is referred to local atmospheric pressure.
RHE Reversing heat exchangers: used to clean up air instead of PPUs for
elimination of and which could block up cryogenic ASUs.
SCF Standard cubic foot: a cubic foot of gas at standard temperature and
pressure (US standard unit of gas).
SIVL Super-insulated vacuum line: pipework for LN consisting of dual pipe
with the annular space filled with evacuated silver-coated Mylar insulation.
Snifting Venting a gas cylinder or gas line momentarily before connection.
This ensures that any debris that has accumulated in the valve outlet is ejected.
It also checks that there is something in the cylinder. It is not recommended,
however, for toxic, reactive or inflammable gases.
Stenching The addition of small amounts of mercaptans or other smelly
compounds to help people smell gas leaks.
TIG welding Tungsten inert gas welding, electric arc welding with a
non-consumable electrode and an inert gas to prevent oxidation of the
tungsten and oxygen in the joint.
Town gas Town gas consists mainly of and CO. It is still in use in
parts of the world where natural gas is expensive and coal is cheap.
Ullage The gas-filled space above the liquid level in a tank.
Vacced cylinder Preevacuated cylinder: a precaution ensuring the
minimum contamination from air ingress.
VIE/VIT Vacuum-insulated evaporator/vacuum-insulated tank: a tank
for cryogenic liquid storage. VIE refers to a high-pressure unit, as these
are used mainly for gaseous supply; VIT refers to a low-pressure unit.
Water gas A flammable mixture of and CO produced by passing
water through hot coke. The reaction is endothermic and heat must be
supplied periodically, typically by interrupting the process while air is
pumped in.
General terminology
Absolute temperature Temperature referred to the coldest temperature
physically possible, normally measured in degrees Kelvin. The Kelvin
scale of absolute temperature has degrees the same size as the Centigrade
scale. For example, 0C, the freezing point of water, is 273.16 K, whereas
absolute zero is 273.16C and 0 K.
Adiabatic A process involving no input or output of energy. For
example, in the adiabatic expansion of a gas, the gas cools because heat
energy has not been supplied from the walls of its container.
Azeotropic mixture A mixture of liquids which when boiled yields a
vapour mixture of the same composition, a constant boiling mixture.
CFCs Chlorofluorcarbons: a special selection of the halocarbons. Most
of these compounds are now implicated in the destruction of stratospheric
ozone and will now not be designed into new equipment or processes.
Most processes will use alternative compounds by 1995 or 2005 under the
Montreal Protocol international directive and national directives.
Critical pressure The pressure above which a gas, no matter how
compressed, will not form a liquid.
Critical temperature The temperatures above which a gas, no matter
how compressed, will not form a liquid.
Differential pressure The difference in pressure between two
points.
Dry ice Solid
APPENDIX A: GLOSSARY 541
Explosive limits (LEL, HEL) The upper and lower percentage byvolume
limits of a flammable mixture of a gas or vapour with air. Below the lower
explosive limit (LEL) the mixture is too weak (has too much air) to ignite.
Above the higher explosive limit (HEL) the mixture is too rich (has too
much fuel) to ignite.
Flashback The propagation of a flame back down a pipe containing a
mixture of fuel and oxidant. This is a dangerous condition, normally avoided
by keeping the fuel and oxidant separate until nearly at the burner and by
fitting flashback arresters. These are typically metal gauze filters which
conduct heat away from a flame in the same manner as the gauze around the
classic Davy miners safety lamp.
Fusible plug Fitted to gas cylinders, these melt at a fairly low temperature
(approximately 150C) and release the contents of a gas cylinder before the
the cylinder walls rupture.
Gas constant This constant has the value (symbol R).
Halocarbon A compound of carbon with one or more of the halogens
fluorine, chlorine, bromine and iodine.
JouleThompson effect The cooling effect seen in all gases below a certain
temperature, when they are expanded at a constriction from a high to a low
pressure.
Latent heat The heat energy absorbed to make a solid melt or sublime, or
make a liquid boil (latent heats of fusion, sublimation and vaporisation
respectively).
Mole molecules.
Montreal protocol The international agreement which controls production
and use of ozone depleting substances such as CFCs.
Noble gas The highly unreactive atomic gases of column 0 of the Periodic
Table of the Elements, i.e. helium, neon, krypton, xenon, radon.
NTP Normal temperature and pressure, i.e. 20C and 1.013 25 bar.
OEL Occupational exposure limit: the concentration of an airborne
substance, averaged over a reference period (8 h, or 15 min for substances
with acute effects), at which, according to current knowledge, there is no
evidence that it is likely to be injurious to employees if they are daily exposed
by inhalation to that concentration (UK).
Pascal The scientific unit of pressure. Rarely used in industry.
Rare gas See noble gas.
Specific heat The heat energy required to raise the temperature of 1 g of
substance through 1 K. All gases have a specific heat of between 2.5 and 3.5
times R, the gas constant. The specific heat at constant volume of a gas
is alway R less than the specific heat at constant pressure
Stoichiometric A stoichiometric mixture of A and B is one in which, when
reaction occurs forming a compound such as AB or or similar, no A or
B is left over.
STP Standard temperature and pressure, i.e. 0C and 1.013 25 bar.
TLV Threshold limit value: the maximum level of a gas to which people
can be exposed on every working day (USA).
Triple point The triple point is the pressure and temperature at which
vapour, liquid and solid can coexist in equilibrium.
Vapour pressure The vapour pressure of a substance is the pressure
exerted, at equilibrium, of vapour in contact with that substance in liquid or
solid form, where the liquid, vapour and container are all at the same
temperature.
Tradenames
Arcton Chlorofluorocarbons (ICI).
Argoshield Argon-based gas mix for MIG welding (BOC).
Astec Helium-based welding gas (Air Products).
Calor An important liquified petroleum gas trademark (Imperial Group).
Coogar Argon-based gas mix for MIG welding (Air Products).
Fomblin Perfluoropolyether oils and greases used with oxygen service and
especially with lubricated vacuum pumps and equipment (Montedison).
Freon The key trademark for chlorofluorocarbons (DuPont).
Helishield Helium-based welding gas (BOC).
Polygas Aerosol propellant system based on a microporous polymer,
acetone and (BOC/Polygas International).
Appendix B: The human nose as a gas detector
Nose anatomy
Inside the nose lies a patch a few square centimetres in area bristling with
sensitive sensor cells. Immersed in a thin layer of mucus lie millions of tiny
moving hairs, cilia, connected to the olfactory cells. These have an electrical
output via their axons that connects them to the nerve cells in the olfactory
bulb, i.e., the region of the brain that interprets the olfactory cells signals
into smells. Some idea of the complexity of the nose can be gained by noting
that there are around 10 million receptor cells.
Nose limitations
The nose is quite limited in its response by its chemical nature and because of
its development to serve animals in the wild, where light gases are in general
unimportant. There is both an upper and a lower size limit for molecules
which can be smelled. The upper limit is set because only those compounds
which have an appreciable vapour pressure at room temperature, which
naturally includes all gases, will reach the nose. The lower limit seems to be
set by the absorption of the compound in the mucus membrane in which the
sensitive cilia are set. Very light gases or those with low solubilities, such as
methane, hydrogen or argon, are undetectable.
Gases which are highly inert, such as the lighter saturated aliphatic
compounds, methane, ethane, propane, butane and pentane, are also
undetectable. Another awkward fact about smell is the great variation in its
efficiency for different compounds. A mercaptan might be detected in parts
per trillion concentration, while ethylene is faint even at a few per cent. A
final limitation to the nose performance is that any gas which is continuously
present is undetectable, i.e., in electronic engineering terminology, the nose
is a.c.-coupled. Gases that are always present in the nose, such as oxygen
and water vapour, are undetectable. Any gas present for a long time will
become less and less sensitively perceived and finally fade almost com-
pletely. The latter is an essential defence mechanism. The brain would
otherwise be overloaded with continuous smell signals in many environ-
ments.
Nose theory
The details of how the nose works are still to a considerable extent mysterious.
A successful theory would have to account for which compounds can be
smelled, the enormous variation in sensitivity to compounds, the large
number of identifiably different smells, and would ideally predict the smell of
new compounds.
Theories which seem to have some viability are as follows.
Specific receptor theory, i.e. Amoores theory. Amoore et al. (1970)

proposed that there are seven different receptors in the nose, each sensitive
only to one class of compounds (Figure A.1), according to molecular shape
and charge distribution. Over the years since this theory was first
presented, however, no such specific receptors have been identified and
many problems have arisen in using the theory.
Membrane puncture theory, i.e. Davies theory. In this theory (Davies and
Taylor, 1954, 1959; Davies, 1965) gas molecules interact with a sensitive
membrane in such a way as to puncture it and allow a flow of sodium and
potassium through, generating a nerve signal directly. Different mem-
brane components will interact differently with different smelling gases.
Again, although attractive as a theory, no sensitive puncturing membranes
have been found, although Davies showed a good correlation between the
catalytic acceleration of saponin haemolysis of red blood cells and the
strength of smell of a range of substances from methanol to musk.
Matrix theory, i.e. Ericksons theory (Engen and Trygg, 1982). There is
some experimental support for this from neurophysiology. If measure-
ments are made at receptor level, there seems to be little predictable
difference between receptor response. However, similar measurements
at the olfactory bulb, after pre-processing of the raw signals by the brain
nerve cells, seem to show more order and selectivity. Basically, this theory
maintains that the receptors are similar but do have sensitivities to
different molecules that are different. These slightly differing receptor
signals are then used as inputs to a matrix, constructed out of neurones,
the outputs of which are specific to a smell. (This technique is sometimes
used to extract gas analysis results from a mass spectrum of mixtures,
section 2.3.6.)
Gas chromatography theory, i.e. Mozells theory (Engen and Trygg,

1982). There is some experimental support from this in the form of time
delays in the response from different parts of the olfactory sensor area in
the nose. Mozell and others have argued fairly successfully that the time
delays inherent in gas molecules diffusing through the mucus to the sensor
cilia are present and could be used as part of the input information in a
matrix type of decoding.
APPENDIX B: THE HUMAN NOSE AS A GAS DETECTOR 545
Figure B.1 Amoore theory of smell receptor sites.
The truth of the matter may be a simple but as yet unthought of theory or it
may be a complex combination of the ideas above. Whatever the truth, there
are some rules of thumb which connect smell to molecular structure.
Moncrieff in his seminal book The Chemical Senses lists 22 general rules and
ennumerates examples connecting chemical make-up with smell over 120
pages. This hardly amounts to a satisfactory theory, however, and even
Moncrieffs 22 rules admit to many exceptions!
Using the nose
The limitations of the nose as a gas sensor are many. Unfortunately, many
industrial gases fall into undetectable categories, at least in their pure state.
It is also unfortunate that oxygen deficiency is not detected. Even amongst
those gases which can be smelled, some are only detectable in concen-
trations at which they are already dangerous. Increasingly, toxic and
flammable gas and oxygen deficiency monitoring is installed as a matter of
course in places regularly handling gases, even where the gas concerned is
easily detectable.
Enhancing the efficacy of the nose is a practice that should perhaps
become more common. The prime example (and the only common
example) is the addition of minute amounts of powerfully unpleasant
smelling thiophane or mercaptans to domestic fuel supplies. (There is
another example from Japan where oxygen gas is sold in small amounts in
aerosol cans and used for inhaling for its refreshing effect. The gas is
normally scented with mint and strawberry essences.)
Is it perhaps too fanciful to suggest that different gases should be coded
with different powerful smells, much as gas cylinders are colour-coded with
paint now? Flammable gases could use mercaptans, simple asphyxiant gases
could use ionone (a powerful violet essence), while toxic gases could be
equipped with a pungent ammoniacal addition, for example.
The mercaptans are so powerful that when a small container of this fuel
gas perfume was accidentally spilled in Manchester, UK, the fire brigade
and gas company offices were flooded with hundreds of calls reporting gas
leaks, some from several kilometres away from the spill. (The human nose is
reported to have a threshold sensitivity of 30 parts per trillion for
ethyl mercaptan,
Fortunately, in many gases which would be otherwise be undetectable
there are adventitious impurities, which, although present at a low level,
give the gas a clear smell (Table A.1). A classic example of this is acetylene,
which often smells of garlic if made from calcium carbide (presumably
because of phosphorus or arsenic impurities) and of acetone if taken from a
cylinder (commercial acetylene is supplied dissolved in acetone). (There is
perhaps a hidden danger here that gas users who come to rely on these
impurities for giving them warning may come unstuck if the process to make
the gas is modernised, giving a purer product, or if the gas comes from a
source other than the usual one and the alternative source gas has no smell.)
Appendix C: Use of industrial gases in schools
Experiment is a key principle of scientific research. In chemistry and physics
at least, experiment should also be a key part of teaching science. There will
always be matters unexplained in textbooks and papers which are only
clarified when an actual trial is carried out.
For educational purposes the preparation of gases by chemical reaction
may often be valuable as it teaches pupils a preparative reaction as well as
the task in hand. This view is a valid one, especially applying to the first time
the gas is used. However, there are many applications for which a
reasonable degree of purity and controllability is needed and these are
better served by an industrial cylinder supply. The preparative reaction can
still be carried out but the cylinder supply can be used for the next stage on
the this is some I prepared earlier principle often seen, perforce, in the
classroom.
Many schools already use industrial gases for a small number of
applications but many more schools could use them for more applications.
These applications could be aimed both at the didactic element of the school
curriculum and at the illustration of current industrial practice.
Industrial gases in compressed cylinders or liquid form are, in many ways,
a very good buy for schools. They provide large quantities of very pure gas,
at a very low price. For example, carbon dioxide is one of the simplest gases
to prepare via a chemical preparation route but some maintenance and
materials costs are involved and the gas contains significant impurities of
water and acid spray, whilst cylinder gas costs around 12 per cubic metre
delivered. Even when cylinder rental/maintenance charges (around 2030
per annum) are taken into account, industrial cylinder gases still represent a
bargain. For gases such as argon and helium chemical preparations are
impossible, whilst for oxygen, ethylene or nitrogen they are relatively
difficult.
Nitrogen
Inerting chemical reactions
There are many reactions commonly avoided in the school chemistry
laboratory that have didactic value but which are rendered possible only
with an inert atmosphere. Similarly, in industry, many important reactions
are often performed under nitrogen atmosphere, which is generally cheaper
to achieve than vacuum conditions.
Dilution
It is frequently necessary to limit reaction rates, for example to avoid
excessive temperature rise, and nitrogen is generally used as a carrier gas to
dilute an active reactive gas. The classic example of this is combustion,
where flame temperature and burning rate are dependent on the nitrogen in
an oxygen/nitrogen mixture. Organic oxidation reactions are possible with
nitrogen dilution of oxygen, which would proceed to complete oxidation
(combustion) otherwise.
Liquid nitrogen
Liquid nitrogen has been traditionally difficult to supply economically in
small quantities and this problem will continue to dog attempts to use LN in
the classroom, at least for smaller schools. Perhaps it would help if someone
(perhaps the gas suppliers) could list universities and other institutes willing
to allow schools to have access to and collect their own supplies from their
own large storage facilities.
Embrittlement of solids
The embrittlement of rubber and plastics is, of course, an important
industrial application. Embrittlement of many materials also makes a
spectacular demonstration and a vivid reminder that properties vary with
temperature.
Freezing of biological systems
It is a comparatively simple matter to rapidly freeze and then thaw soft fruit,
such as raspberries, and compare with the product thawed after a slow freeze
in a standard refrigerator. The industrial importance of this needs no
emphasis here.
It should also be possible to design suspended animation experiments
with yeast or even small insects.
Cooling of YBCO superconductors
Many schools with keen chemistry departments managed, as early as
1988, to make a crude sintered YBCO pellet, which when cooled to LN
APPENDIX C: USE OF INDUSTRIAL GASES IN SCHOOLS 549
temperatures, attains the uniquely interesting state of supercon-
ductivity.
82
Oxygen
The justification for the use of oxygen, rather than air, in most applications
revolves around the intensification of the chemical reaction achieved with
oxygen.
Oxy-gas flames
There are many operations for which an oxy-gas flame can be used to
advantage in the school laboratory. For example, many more metals can be
melted in the methane/oxygen flame than with a methane/air flame. The
ultimate, of course, is the 3100C oxy acetylene flame.
Oxygen in chemical reactions
There are a number of reactions in which the regulation of oxygen
concentration can be demonstrated to showdependence of reaction rate and
thus rate equations.
Argon and helium
Sintering
Many of the achievements of powder metallurgy have been made possible
by the advent of pure inert gases. A demonstration of sintering technology
should be possible on a small scale using apparatus already available in the
laboratory.
Speed of sound
Simple whistles or organ pipes easily demonstrate the dependence of the
speed of sound on molecular weight.
82
The author tested a number of pellets of the higher current superconductor BISSCO as
part of an industrial materials development programme. It was found that with the simplest of
apparatus many hundreds of amps could be passed through samples of less than using
simple welding power supplies.
Argon arcs and plasma
The requirement for a powerful electrical source, typically at RF frequen-
cies in the case of plasma, restricts school experiments with these interesting
phenomena.
Welding
An important part of metallurgy studies, frequently undertaken by the
chemistry department in schools, should be the study of welding. The
preparation of TIG/MIG welds, and their assessment relative to the gas used
argon mixtures) and relative to stick electrodes, is rarely undertaken,
however, despite the fact that mini-MIG welders are now available for
around 100. The involvement of the physics department, observing arc
temperature and electrical arc characteristics, could also be useful.
Helium buoyancy
The contrast between the transport properties (i.e. diffusivity, viscosity
and thermal conductivity) of a very light gas such as helium and those of air
are marked and could be employed, for example, in apparatus for the
measurement of mixing, damping of mechanical oscillation and double-
glazing performance.
Hydrogen
A simple gas chromatograph with a flame ionisation detector can easily be
built for a couple of hundred pounds in school; the parts are readily available
and not expensive.
83
Today, with an inexpensive PC-based data system,
using a computer that the school will already have, a fairly sophisticated GC
is possible. Such a system is easily capable of separating mixtures of organic
compounds and is a useful adjunct to the study of organic chemistry, helping
with the determination and understanding of side-reactions. Safety aspects
are mostly taken care of by attention to leaks and ensuring that the hydrogen
is shut off should the FID detector flame fail for any reason.
Fat hardening
Fat hardening is a very important industrial process and could probably be
modelled in the school laboratory. The use of finely divided industrial
83
The author built one whilst at school. It comprised a commercially available packed column
with a home-made heating bath and a home-made FID/electrometer, running off a cylinder of
hydrogen, with the electrometer fed to a chart recorder. The most expensive item turned out to
be the chart recorder.
APPENDIX C: USE OF INDUSTRIAL GASES IN SCHOOLS 551
catalysts and pure grades of oil enables fairly safe low temperatures to be
used. Obviously, the combination of heat and hydrogen needs careful
attention to safety.
Carbon dioxide
Dry ice can be used in schools for cryogenics and also for gas lubrication.
The unique sublimation of at atmospheric pressure has been used for
years in the food industry as a uniquely simple freezer/refrigerator. Less
well-known are applications in which a thin film of derived from dry ice
pellets is used to provide a high-pressure gas bearing.
Physics courses have for years used ice pucks for kinematics experi-
ments. These are actually metal pucks using an almost frictionless
bearing to glide frictionlessly across a glass-topped table. The bearing is
simply a hollow underneath the puck in which a small piece of dry ice can be
placed. As the puck slides along it warms the dry ice, producing gas
which escapes around the smooth skirt of the puck, raising it a fraction of a
millimetre above the glass table.
The cryogenic properties of dry ice can be exploited in many ways. It can
be used, for example, in cooling methanol used in making a diffusion
(continuous) cloud chamber capable of revealing nuclear particle tracks.
Gaseous can be used in respiration demonstrations. The respiration
rate (in breaths per minute and also in terms of production) of creatures
depends on concentration. Demonstrating these important physio-
logical effects may require the use of a gas analyser. However, as noted
above, a crude gas analyser can be quite simple or, alternatively, a
straightforward wet chemical technique could be used.
It should be possible to show the effects of a gas layer on the heating
and cooling of a black body under the sun or a tungsten lamp. This is the
basis of the greenhouse effect, which is now a much-discussed environ-
mental issue.
Ethylene and propylene
Ethylene is now important in the school chemistry syllabus, with its
emphasis on organic chemistry, and is relatively difficult to make from liquid
chemicals. Ethylene is available in relatively high purity at low cost in bulk
from industrial gas companies. However, in most countries it is currently a
relatively expensive special gas in small quantities because it is used by only a
few small users owing to the superiority of acetylene in oxy-fuel and a lack of
other small-scale industrial uses. Nevertheless, if a sizeable number of
schools began to purchase it, then prices would tumble and it could usefully
be provided to schools for classic organic chemistry experiments.
Propylene is a suitable substitute in many instances. It is much more
readily available at reasonable cost and can be used to illustrate ethylene-
based reactions.
Acetylene
The use of acetylene in flames is obvious. Less obvious, but interesting and
still of some industrial importance, are the chemical reactions of acetylene.
Acetylene, because of its reactive carboncarbon triple bond, will react
relatively easily in many saturation reactions.
Appendix D: Gas cylinder colour marking
Different gas companies use different colour schemes around the world,
depending on their own standards and local regulations. The colours also
differ for cylinders intended for medical use. All gas users should always
ensure that they are fully informed on the colours used in their country by
their own suppliers.
There are also different schemes around the world for gas cylinder
connections. Connections differ in screw-thread pitch, direction and
diameter. They also differ in the diameter and form of the gas sealing
surfaces and the use of sealing washers. In total there are hundreds of
different connection types and this helps to prevent incorrect gases being
connected. As a rule, oxidising gases, such as air or oxygen, use a normal
(right-hand) thread, whilst flammable fuel gases use a left-hand thread.
In Europe, DIN (Deutsches Institut fr Normung, Germany) or BCGA
(British Compressed Gas Association, UK) standard connections are
employed, whilst in the USA the CGA (Compressed Gas Association)
connections are used. Standard bodies, such as CGA and BCGA, issue
up-to-date information on connections, as do the gas suppliers themselves.
References
Agricola (1950, orig. 1556) De Re Metallica, Hoover Publications, New York.
Amoore, J.E., Johnston, J.W. Jnr and Rubin, M. (1964) The stereochemical theory of odor.
Scientific American, 210(2), 4249.
Andeen, C. (1971) A capacitive gauge for accurate measurement of high pressure. Rev. Sci.
Instr., 42, 495496.
Anon (1975) High Pressure Engineering, Proceedings of the 2nd International Conference,
Mechanical Engineering Publications (Inst. Mech. Eng.), London.
Anon (1980) Priestley in London, Royal Society of Chemistry, London.
Anon (1989) Separation of Gases, 5th BOC Priestley Conference, Proceedings of the Royal
Society of Chemistry, London,
Anon (1993) The Role of Oxygen in Improving Chemical Processes, 6th BOC Priestley
Conference, Proceedings of the Royal Society of Chemistry, London.
Anon (1995) Happy birthday hel i um, New Scientist 146(1972), 8th April.
Aykroyd, W.R. (1935) Three Philosophers, Heinemann, London.
Berbesson, J. (1979) Long distance transport of LNG via cryogenic pipeline, in New
Technologies for Exploration and Exploitation of Oil and Gas Resources, Vol 1, Graham &
Trotman, London, p. 581.
Bridgman, H.A. (1990) Global Air Pollution: Problems for the 1990s, Belhaven Press.
London.
Buchanan, R.A. The Power of the Machine, Viking, London.
Calder, A. (ed.) (1995) Technology Foresight Chemicals, HMSO, London.
Carr, P. (1994) In situ gas analysis. European Semiconductor, Sept., 13.
Cherin, A.H. (1983) Introduction to Optical Fibres, McGraw-Hill, Tokyo.
Collins, M. (1988) Liftoff, Grove Press, New York.
Corbett, M.J. and Smith, R.B. (1993) in The Role of Oxygen in Improving Chemical
Processes, Proceedings of the 6th BOC Priestley Conference, Royal Society of Chemistry,
London.
Coulson, J.M. and Richardson, J.F. (1970) Chemical Engineering, Pergamon, Oxford.
Davies, A.C. (1965) Journal of Theoretical Biology, 8, 1.
Davies, A.C. and Taylor, F.H. (1954) Nature, 174, 693.
Davies, A.C. and Taylor, F.H. (1959) Biology Bulletin, 117(2), 222.
Dennis, N.T.M. and Heppell, T.A. (1968) Vacuum System Design, Chapman & Hall,
London.
Denny, R. (1977), in Oxygen in the Metal and Gaseous Fuel Industries, Proceedings of the 1st
BOC Priestley Conference, Royal Society of Chemistry, London, pp. 232255.
Downie, N.A. (1991) Cryogenic system for fault current l i mi t er. Research Disclosure, 330,
829.
Dunbobbin, B.R. (1987) MOLTOX process. Gas Separation and Purification, 1, 23.
Einstein, A. (1924) Z. Electrochemie 14, 235.
Elvin, L. (1971) Organ Blowing: its history and development, Laurence Elvin, Lincoln.
Enge, H. (1966) Introduction to Nuclear Physics, Addison-Wesley, London.
Engen, T. (1982) The Perception of Odors, Academic Press, New York, pp. 1734.
Fricke, J. (1988) Aerogels. Scientific American, May, 9297.
Gibbs, F.W. (1965) Joseph Priestley: Adventurer in Science and Champion of Truth, Thomas
Nelson, London.
Gibbs, F.W. (1967) Joseph Priestley, Doubleday & Co.
Haldane, J.B.S. (1927) On being the right size, in Possible Worlds, Chatto & Windus,
London.
Hendrick, D.J. and Sizer, K.E. (1992) Breathing coal mines and surface asphyxiation from
stythe (black damp). British Medical Journal, 305, 509.
Hands, B. (1986) Cryogenic Engineering, Academic Press/Harcourt Brace, London.
Harper, G.S. and Parrish, G. (1985) Production Gas Carburising, Pergamon Press, Oxford.
Hirose, T. (1987) Thermal Production of Diamond Thin-films, Proceedings of the Osaka Sanso
Conference, Tokyo. Osaka Sanso, Osaka, Japan.
Hrubesh, L.W. (1987) Optical Characterisation of Silica Aerogel Glass, Lawrence Livermore
Laboratory, July report.
Hunt , J. (1954) The Conquest of Everest, Hodder & Stoughton, London.
Irani, C.A. and Funk, E.W. (1977) Separations using supercritical gases, in Recent Develop-
ments in Separation Science, CRC Press, Florida, pp. 171193.
Isalski, W.H. (1989) Separation of Gases, Oxford University Press, Oxford.
Jolley, S. (1990) European Patent No. 0487183, May.
Jones, T. (1988) Corporate Killing: Bhopals will happen, Free Association Books, London.
Kapteijn, F. and Singoredjo, L. (1993) Catalyst and catalytic reduction. World Patent No.
9302776.
Kaye, B. (1995) Conservation of waterlogged archaeological wood. Chem. Soc. Rev., 24, 35.
Kirk-Othmer (1978) Encyclopaedia of Chemical Technology, 3rd edn, Wiley 1982.
Lavoisier, A.L. (1952) Elements of Chemistry, Encyclopaedia Brittannica/William Benton,
Chicago.
Melvin, A. (1988) Natural Gas, Adam Hilger, Bristol.
Monnot, G. (1985) Principles of Turbulent Fired Heat, Gulf Publishing Co/Editions Technip
Paris (original in French).
Mooney, M.M. (1973) The Hindenburg, Hart-Davis MacGibbon, London.
Ohmi , T., Okumura, T., Sugiyama, K., Nakahara, F. and Murota, J. (1988) Outgas-free
corrosion-resistant surface passivation of stainless steel for advanced ULSI processing
equipment. Extended Abstracts: 174th Electrochemical Society, Fall Meeting, Paper 396,
pp. 579580. Electrochemical Society.
Perkins, D.H.P. (1972) Introduction to High Energy Physics, Addison-Wesley, London.
Reid, R.C. (1977) The Properties of Gases and Liquids, McGraw-Hill.
Wightman, W.P.D. (1950) The Growth of Scientific Ideas, Oliver and Boyd, Edinburgh.
Williams, T.I. (1981) A History of the British Gas Industry, Oxford University Press, Oxford.
Yang, R.T. (1987) Gas Separation by Adsorption Processes, Butterworths, London.
Zuwala, M.S., Craig,G.R. and Cass, D.L. (1994) Cryogenic processing of used tyres and other
materials. World Patent No: 9407670.
Further reading
Ahlberg, K. (1985) AGA Gases Handbook, AGA AB, Lidingo, Sweden.
Alexander, W. and Street, A. (1982) Metals in the Service of Man 8th edn, Penguin,
Harmondsworth, UK.
Anon (1977) Oxygen in the Metal and Gaseous Fuel Industries, 1st BOC Priestley Conference,
Proceedings of the Royal Society of Chemistry, London.
Anon (1978) Basic Oxygen Steelmaking: a new technology emerges?, Proceedings of The Metals
Society Conference, May 45, The Metals Society, London.
Anon (1987) Distillation and Absorption 1987, Institution of Chemical Engineers, Rugby.
Anon (1990) STEP 1990, Proceedings of SEMI Conference, Brussels. SEMI, Mountain View,
California.
Anon (1991) RADTECH Europe 1991, RADTECH International, Northbrook, Illinois.
Anon (1995) CRC Handbook of Chemistry and Physics, 76th edn, CRC Press Inc, Boca Raton,
Florida.
Atkinson, H.V. and Rickinson, B.A. (1991) Hot Isostatic Processing, Adam Hilger, Bristol.
Banks, R.E. (1980) Fluorocarbons and their Derivatives, 2nd edn, MacDonald Technical,
London.
Barnard, J.A. and Bradley, J.N. (1995) Flame and Combustion, 3rd edn, Chapman & Hall,
London.
Belsterling, C.A. (1971) Fluidic Systems Design, Wiley-Interscience, New York.
Board, K. and Morgan, D.V. (1985) Semiconductor Microtechnology, Wiley, Chichester.
Brauer, H. and Varma, Y.B.G. (1981) Air Pollution Control Equipment, Springer-Verlag,
Berlin.
Braithwaite, A. and Smith, F.J. (1995) Chromatographic Methods, 3rd edn, Chapman & Hall,
London.
Butrica, A. (1980) Out of Thin Air, Praeger Publishers Inc., Westport, Connecticut.
Cayless, M. A. and Marsden A.M. (eds) (1983) Lamps and Lighting, Edward Arnold, London.
Cherin, A.H. (1983) Introduction to Optical Fibres, McGraw-Hill,
Clarke, K.P., Coulson, D., Gibbs, E.D., Keon, K.B. and Kruijshoop, A.F. Optical Fibre
System Telstra Corporation patent WO 9422241.
Cowper, C.J. and DeRose, A.J. (1985) The Analysis of Gases by Chromatography, Pergamon
Press, Oxford.
Davidson, F.H. and Harrison, H. (1963) Fluidised Particles, Cambridge University Press,
Cambridge.
Davies, A.C. (1992) The Science and Practice of Welding, 10th edn, Cambridge University
Press, Cambridge.
Desrosier, N.W. and Tressler, D.K. (1977) Fundamentals of Food Freezing, AVI Publishing
Co. Inc., Westport, Connecticut.
Douglas, J.F., Gasiorek, J.M. and Swaffield, J.A. (1984) Fluid Mechanics, Pitman Publishing,
London.
Dumas, T. and Bulani, W. (1974) Oxidation of Petrochemicals, Applied Science Publishing,
London.
Eilbeck, W.J. and Mattock, G. (1987) Chemical Processes in Waste Water Treatment,
Wiley/Ellis Horwood, Chichester.
Fifield, F.W. and Kealey, D. (1983) Analytical Chemistry, Chapman & Hall, London.
Flowers, B.H. and Mendoza, E. (1970) Properties of Matter, John Wiley, Chichester.
Fogg, P.G.T. and Gerrard, W. (1991) Solubility of Gases in Liquids, John Wiley, Chichester.
Fraas, A.P. and Ozisik, M.N. (1965) Heat Exchanger Design, Wiley, Chichester.
Gordon, J.E. (1981) Structures or Why Things Dont Fall Down, Penguin, London.
Hausner, H.H. (ed.) (1971) Modern Developments in Powder Metallurgy, Plenum, New
York.
Heaton, C. A. (ed.) (1984) Introduction to Industrial Chemistry, Chapman & Hall, London.
Houghton, J.T., Taylor, F.W. and Rogers, C.D. (1986) Remote Sounding of Atmospheres,
Cambridge University Press, Cambridge.
Hutzinger, O. (1980) The Handbook of Environmental Chemistry, Springer-Verlag, Berlin.
Kaye G.W.C. and Laby T.H. (1995) Tables of Physical and Chemical Constants, 16th edn,
Longman, London.
Kletz, T. (1992) HAZOP and HAZAN, Institution of Chemical Engineers/Hemisphere,
Rugby.
Kohl, A.L. and Riesenfeld, F.C. (1960) Gas Purification, McGraw-Hill, New York.
Lacey, R.E. and Loeb, S. (1972) Industrial Processing with Membranes, Wiley-Interscience,
New York.
Ledwith, A. and Moss, S.J. (1987) The Chemistry of the Semiconductor Industry, Blackie,
Glasgow.
Linan, A. and Williams, F.A. (1993) Fundamental Aspects of Combustion, Oxford University
Press, Oxford.
Liu, B.Y.H. (ed.) (1976) Symposium on Fine Particles, Academic Press, New York.
McMillan, F. et al. Treatment of Scrap, Patent WO 8002672, Commonwealth Industrial Gases.
Meyer, E. (1977) Chemistry of Hazardous Materials, Prentice-Hall, New Jersey.
Moncrieff, R.W. (1967) The Chemical Senses, 3rd edn, Leonard Hill, London.
Moore, W.J. (1978) Physical Chemistry, Longman, London.
Morgan, D.V. and Board, K. (1985) Semiconductor Microtechnology, John Wiley, Chichester.
Mosely, P.T., Norris, J.O.W. and Williams, D.E. (eds)(199l) Techniques and Mechanisms in
Gas Sensing, Adam Hilger, Bristol.
Nash, P. (1983) The Essentials of Foam, Dry Powder, and Gaseous Extinguishing Systems,
Building Research Establishment, Borehamwood.
Newstubb, P. F. (1969) Mass Spectrometry and IonMolecule Reactions, Cambridge University
Press, Cambridge.
Nunes, A.C. (1977) Gas welding origins. Welding Journal, June, 56, 1523.
Pasek, A.D. (1991) Pool Boiling on Porous Surfaces, PhD Thesis, Cryogenics Institute,
University of Southampton.
Patterson, H.B.W. (1983) Hydrogenation of Fats and Oils, Applied Science Publishers,
London.
Perry, R.H. (1984) Perrys Chemical Engineers Handbook, 6th edn, McGraw-Hill, Maiden-
head, UK.
Pichot, P. (1986) Compressor Application Engineering, Gulf Publishing, Houston.
Roberts, T.A. and Skinner, F.A. (eds) (1983) Food Microbiology: Advances and Prospects,
Academic Press, London.
Ruelle, D. (1991) Chance and Chaos, Princeton University Press, Princeton.
Sale, F. (1979) Steel Production, Open University Press, Milton Keynes.
Scott, R.W.W. (ed.) (1982) Developments in Flow Measurement, Applied Science Publishers,
London.
Scurlock, R. (1992) History of Cryogenics, Clarendon Press, Oxford.
Spitzer, D.W. (1990) Industrial Flow Measurement, Instrument Society of America, North
Carolina.
Strauss, W. (1966) Industrial Gas Cleaning, Pergamon Press, Oxford.
Thelning, K.E. (1975) Steel and its Heat Treatment, Butterworths, London.
Thorne, S. (1983) Developments in Food Preservation 2, Applied Science Publishers,
London.
Walker, M.J. (ed.) (1993) Concrete in Hot Climates, Chapman & Hall, London.
Walker, R.D. (1983) Small-scale steelmaking, Applied Science Publishers, London.
Yang, R.T. (1987) Gas Separation by Adsorption Processes, Butterworths, London.
Yinon, J. and Shmuel, Z. (1981) Analysis of Explosives, Pergamon Press, Oxford.
Zemanski, M.W. (1968) Heat and Thermodynamics, McGraw-Hill, New York.
Index
Accidents 26388
leaks 263
oxidant 2748
reactive gases 2789
toxic gases 2635
Acetone 122
Acetylene 39
absorbers 122
cutting 31719
diamond production 533
future 525
lamp 5
in liquid oxygen 84
production from calcium carbide 121
production from hydrocarbons 121
school chemistry 5512
thermal blasting of stone 474
welding 3245
Adduct purification 421
Adiabatic compression ignition 267, 2768
Adiabatic cooling 745
Advanced Silicon Materials Inc 136
Aerosol can, see Spray can
AFROX Company 308
AGA Company 45
Air Products Company 45, 522
Air quality monitoring 18891
chemical tape monitors 189
electrochemical cells 18990
flammable gases 1889
heated sample lines 190
human sense of smell 191, 543
membrane MS 1901
multipoint MS 190
oxygen depletion 188
photoionisation 190
Taguchi semiconductor bead 1889
AIRCO Company 45, 50, 464, 465
Airships, see Balloons
Alumina, activated 99
Ammonia 477
agriculture 366
balloons 494
cracking for forming gas 1256
production plants 3445
Amoore theory of smell 544
Anaesthesia 5049
AOD, see Argon oxygen decarburisation
APIMS, see Mass spectrometer,
atmospheric pressure ionisation
Arc coating, see plasma coating
Arc cutting 3256
Arc furnace 121, 3056
ARC Machines Corp. 194
Arc welding 3269
fumes 328
MIG 3267
orbital welder 194, 327
TIG 326
Archaeological preservation 5278
Argon 38
argon-41 487
argon credits 52
argon ion laser 492
avoiding dross in casting 311
double glazing 4823
extraction from ammonia plants 3445
filament lamp 4878
fluorescent lamp 48990
liquid argon nuclear particle detector 517
nuclear particle detectors 51617
sputtering 4001, 4802
stirring of molten steel 306
versus vacuum 315
see also Arc welding
Argon oxygen decarburisation 3067
Arsine l35, 254, 390, 420
ASME (USA standards) 196
Atmosphere, composition 84
Autofrettage 244
Avogadros number 2, 11
Balloons 4926
Barrier polymers 114
BCF, see Bromochlorodifluoromethane
Bergbau Forschung Corp. 111
Bernoullis equation 13
Bessemer steel process 304
Bhopal factor 44
Big Three Corp. 50
Biochemical oxygen demand 430, 432, 435
Birdseye, Clarence 351
Black heart in ceramics 4834
Blanketing of tanks 3457
Blast furnace 298301
Blow-down of gas cylinders 251
BOC Group plc 44, 89, 343, 365
BOD, see Biochemical oxygen demand
Boiling 29, 4446
enhanced boiling surfaces 445
film boiling 444
Boron carbide 534
Boron trichloride 129, 135, 423, 405
BOS, see Steelmaking, basic oxygen
processes
Boyle, Robert 1
Breathing 5024
aviation 51314
carbon dioxide in 367, 506
diving 51213
mountaineering 514
Breathing of mine workings 263
Breathing of tanks 235
Bremstrecke, see Filters, Bremstrecke
Brin process 4
British Antarctic Survey 288
British Technology Group plc 477
Brittania Refined Metals Company 331
Broeder Mueller Company 247
Bromochlorodifluoromethane 474
Brownian motion 214
BS (British Standards) 196
Bubble chamber 30
Bubble point 29
Bubbles 29, 4334
Bulk gases 49
Burn-boxes 296, 412
Burners 310, 476, 495
Bursting disc 230, 2823
Butane 1245
Calcium carbide 121
Cambridge Fluid Systems Ltd 384
Can rigidification 364
Carbon, activated 99100
Carbon dioxide 39
agriculture 3656
breathing 506
cleaning 411
coolant in nuclear reactors 487
drinking water 437
environmental cabinets 4501
extraction from ammonia plants 120, 344
extraction from smokestacks 119
extraction from waste streams 49
fire extinguishers 4745
future supplies 525
greenhouse gases 2913
laser 491
oil wells 342, 526
pesticides 3667
spray cans 499
storage 2512
sublimation 32
swimming pools 439
tobacco fluffing 371
tyres and inflatables 4756
see also Dry ice; Supercritical fluid,
extraction
Carbon fibre 5334
Carbon monoxide 125, 137, 291, 300, 378,
509
Carbonated drinks 3624
Catalytic combustion 267
Cavendish, Henry 2
CEA-LETI France 531
Centrifugal gas separation 117
Ceodeux Company 247
Ceramic firing 4835
Cerenkov radiation detectors 517
CERN R&D institute 517
CFC, see Chlorofluorocarbons
Channelling
in distillation 91
in PSAs 105
in purifiers 143
Charpak 517
Chemical vapour deposition 3936,
47982, 5324
Chlorine 126, 129, 411, 436, 439
Chlorine dioxide 431, 441
Chlorofluorocarbons
aerosol spray cans 496
global warming 2923
nuclear particle detectors 516
ozone depletion 28890
refrigerants 471
semiconductor industry 41113
Choked orifice, see critical orifice
Clathrates 100, 122
Claude, Georges 5
Claus process for sulphur recovery
3434
Claw pump, see Northey pump
Cloud chamber 30
Coal gas 3
Coal gasification 3401
Coke oven gas 125
Cold traps 451
Combustion
future use of oxygen 521
see also Flame
Commonwealth Industrial Gases Company
367
Composite lamination 427
Compound semiconductors 393, 4202
Compressibility (Z) factor 9
Compressible flow 1519
Compression fittings 1956
see also Swagelok; VCR
INDEX 561
Compressors 6473
centrifugal 657
cooling 713
isothermal ideal 72
reciprocating 70
screw701
surging 6970
Computational fluid dynamics 1921
Concrete cooling 4589
Conductance 25
Contracts for gas supply 53
Copper extraction 307
COREX process 3023
Corporate structure 44
Cracking of hydrocarbons 33940
Critical opalescence 34
Critical orifice 15, 205
Critical point 324
CRT manufacture 427
Cryogen storage tanks 22931
cluster tanks 234
pressure raising circuit 231
roll-over phenomenon 2345
Cryogenic insulation 2259
aerogel 2278
perlite 227
radiation losses 228
thermal conductivities 229
vermiculite 227
see also Vacuum insulation
Cryogenics 44272
accurate dispensing 4645
biological storage 51011
boiling 4446
cryogenic pipelines 4712
cryosurgery 510
dust-mite treatment 511
embrittlement 4489
enhanced boiling surfaces 4456
freeze-branding 510
gaps in cryogen spectrum 471
phase separator 450
properties of cryogens 443
turbulent flow 449
Cryopumps 452
Cryosurgery 510
CVD, see Chemical vapour deposition
Cyanogen 320
Cyclones 2178
Cyclopropane 505
Czochralski process for crystals 392
Daltons law of partial pressures 9, 506
Darcys rule 18
Decaffeinated coffee 370
Decant filling of gas cylinders 250
Deflagration 274
Deflashing 458
Degreasing oxygen components 275
DENAL Company 136
Deoxo, see Purifiers, catalytic
see also Purifiers, getter
Desflurane, see Fluranes
DESY R&D institute 517
Detarred tobacco 371
Detonation 274
Detonation gun 324
Deuterium 137, 421
Dew point 29
Dewar 27
see also Vacuum insulation
Diamond 5323
Diborane 397, 472
deuterated 421
Diethylether 5067
Differential embrittlement 462
Diffusion 224
back-diffusion 222, 387
in oxide formation 398
Dimethyl ether 499
Dimethylformamide 122
DIN (German Standards) 196
Dinitrogen tetroxide 260, 290, 271, 473
see also Nitrogen dioxide
Disilane l36, 421
Distillation column 912
Distillation of air 8398
argon 957
composition diagram 867
distillation 8592
hydrocarbon absorbers 93
Linde double column 8991
noble gases 978
pre-purification 84
reflux ratio 87
single column plants 889
Diving gases 51213
Double glazing 4823
DOW Chemical Corp. 115
Dross formation
in metal casting 311
in soldering 4256
Dry etching 401
see also Plasma etching
Dry ice 4489
bead-blasting 4678
flash losses 4489
food freezing 35960
sliding pucks 551
theatrical fog 4667
Dry pump, see Northey pump
DuPont Corp. 390
Ecopoints 294
ECR, see Electron cyclotron resonance
Edwards High Vacuum Ltd 412
EFFSET process 311
Electrolysis 1267
Electron beam coating 482
Electron capture detector 171, 224
Electron cyclotron resonance 403
Electronic flow controls 2201
Electrostatic precipitation 218
Emergent economies 59
Endothermic generators 31516
Energy in compressed gas 7
Enflurane, see Fluranes
Entonox, see Nitrous oxide, mixture with
oxygen
Environmental cabinets 4501
Environmental issues 28894
BCF fire extinguishants 474
carbon monoxide 291
CFCs 28890
environmental audit 2934
future progress 534
global warming 2913
greenhouse effect 291
ground level ozone production 2901
Montreal Convention 412
NO
x
290, 4778
semiconductor industry 41113
stratospheric ozone 28892
see also Scrubbers
Epichem Ltd 421
Erickson theory of smell 544
Ethylene 401
production 123
ripening of fruit 3645
school chemistry 551
Ethylene oxide 511
Excess flow shut-off, see Regulators
Excimer 492
Expansion turbines 746
Experience curve 567
Explosions 2734,
Extrusion 4667, see Plastic processing
FAR/fatal accident rate 2878
Fault current l i mi t er 5312
Fermenters, gas supplies/control 535
Fermilab 517
Ferrites 484
Ficks law 22
FID, see Gas chromatograph, flame
ionisation detector
Filament lamp 4878
Filters 21317
air filters for cleanrooms 216
Bremstrecke 214
ceramic 3878
heated, for ultrapure gas 387
inertial impaction 21415
materials 21617
metal 3878
most penetrating particle size 21415
Fingerprinting in gas analysis 520
Fire piston 267
Flame 2679
diffusion 267, 268
flame speed 268, 316
flat flame 477
hardening 319
high emissivity 477
high-velocity 269
premixed 2679
spraying of powders 3234
stabilisation 269
temperatures 31920
turbulent 269
welding 322, 324
Flame cutting 31719
Flammability 26673
BLEVEs 273
explosive limits 266
mixtures 270
oxygen index 2712
Flash losses 4468,
Flashback arrestors 205, 213, 338, 447
Float glass 478
FLOSTAT, see Regulators
Flow measurement 14759
correlation 152
gravimetric 1478
level gauges 1579
liquid cryogens 1568
positive displacement 1501
rotameter 14950
thermal anemometers 1535
thermal mass flow 152
tracer 156
turbine 151
ultrasonic 155
venturi/orifice 1489
vortex shedding 152
Flow restrictors 389
Fluffing of tobacco 371
Fluidic controls 2212
FLUOR Corp. 116
Fluorescent lamp 48990
Fluorine 129130, 492
Fluranes 5067
Food preservatives 3601
Forming gas 125, 420, 478
Freezing food 35160
air-blast 352
belt freezer 355
crust-freezing 356
cryogenic air 3534
dry-ice freezing 35960
immersion freezing 357
in food transportation 357
individual quick frozen 351
liquid nitrogen 3523, 3548
non-stick freezing 3589
spiral freezer 356
INDEX 563
vacuum packing 356
Freeze-branding 510
Freeze-drying 4634
Freeze-grinding 357
Friable buildings 136, 279
Fruit transportation 3645
Fuel cells 523
Fullerenes 123
see also Clathrates
Furnace atmospheres 31116, 3336, 484
Gardner Cryogenics Corp. 48
Gas analysis 16391
anaesthesia 5078
bubbler 1634, 185
chemiluminescence oxygen meter 186
electrochemical cell 1812, 18990
fingerprinting 520
infrared analysers 1725
katharometer 163
magnetic oxygen meter 1867
mercury reduction analyser 186
non-dispersive IR analysers 1734
optical absorption 1878
optical refractometer 176
Raman spectroscopy 175
resistivity 163
silane layer resistivity 1856
zirconia oxygen meter 182
see also Gas chromatograph; Mass
spectrometer; Moisture, measurement
Gas ballast for vacuum pumps 2612, 414
Gas cabinets 21920
Gas chromatograph 16472
back-flushing 168
carrier gases 51516
columns 1667
electron capture detector 171
flame ionisation detector 16971
heart-cutting 168
katharometer 169
school instrument 550
ultrasonic detector 1712
Gas compatibility 20911
Gas controls panels 21819
Gas cylinders 23651
300 bar rating 247
absorber filled 2534, 524
aluminium 243
autofrettage 244
coffin 245
colour 553
colour schemes 553
dip-tube 246
dual port valve 247
empty 2845
filling 24851
gas-to-packaging ratio 237, 242
goose-neck eductor 246
hoop stress 237
internal finish 241
manufacture 238
plethysmography 240
pre-evacuation 251
safety testing, see Pressure safety
testing
snifting 246
spherical 245
stainless steel 242
ultraclean 391
ultracylinders 243
valves 241, 2456
wrapped cylinders 2435, 524
Gas losses 2556
Gas separation 61147
perfect separation 613
Gas storage siting 233, 284
Gas-bags 236
Gaseous chemicals 42
Gasholder/gasometer 2356
Gay-Lussac 2
GC, see Gas chromatograph
Gear pump 71
GEC Semiconductors Ltd 530
Geiger counter 516
GiffordMcMahon refrigerator 452
Glass manufacture 47682
coating processes 47882
low emissivity glass 479
Global warming potential 2913
GMAW, see Arc welding, MIG
GOLDOX process for gold leaching 308
Graphil 534
Grashof number 18
Greenhouse effect 291
Grinding, cryogenic 357
Ground-freezing 456
GTAW, see Arc welding, TIG
Guinness beer, 364
GWP, see Global warming potential
Habbakuk project 443
Haemoglobin 36, 502, 508
Haldane, J.B.S. 226
Hale Hamilton Ltd 383
Halothane 5067
Hard-facing 322
HAZ, see Heat affected zone
HAZAN 2878
HAZOP 2867
HCFC, see Hydrochlorofluorocarbons
Heat affected zone 31819
Heat capacities 11
Heat exchangers 7883
counter-current 7982
co-current 80, 82
effective T 80
refluxing exchangers 92
Heat treatment of metals 31116
annealing 31213
carburising 31415
decarburising 313
nitriding 315
Helium 41920
balloons 4923
extraction 1323
He/oxygen breathing mixtures 5034,
51213
liquefaction 1334
lung function tests 50910
storage 229
Helium-neon laser 4912
Henrys law 35
HEPA, see Filters, air filters for
cleanrooms
High T
c
superconductors 3356, 5312
Hildebrand 3
Hindenburg factor 494, 523
HIP, see Hot isostatic pressing
Hoechst 50
HoechstWacker process for acetaldehyde
343
Hot isostatic pressing 3368
HWT Company 253
Hybrid circuits 41819
Hydrides 253
Hydrochlorofluorocarbons 419
global warming 2923
ozone depletion 28890
Hydrogen 40, 429
ammonia cracking 1256
balloons 4923
ceramic firing 4845
coke oven gas 125
cooling alternators 485
FID 16971
hardening oils 36870
Hindenburg factor 494, 523
hydrogen-2, see Deuterium
invisible flame 321
low ignition energy 271, 321
mercury removal 297
methanol cracking 125
oxy-hydrogen flame 424
PSA 107
reformer gas 125
see also Liquid hydrogen
Hydrogen chloride 130, 492
Hydrogen embrittlement 201
Hydrogen fluoride 130
Hydrogen Supplies Ltd 50
Hydrometallurgy 3078
Hypergolic compounds 266, 271, 472
HYT Corp. 407
IBM 393, 407
ICI 50, 534
Ideal gas equation 7
IDLH, see Maximum levels of toxic/
asphyxiant gases in air
Ignition energy 267, 271
Ignition temperature 267
hydrogen 271
IMEC R&D institute 385
INCAM 386
Inert gas generator 11819, 484
Inerting of tanks, see Blanketing of tanks
Infrared analysers, see Gas analysis,
non-dispersive IR
Institute for Aviation Medicine 514
Intrinsically safe electrics 272
IQF, see Freezing food, individual quick
frozen
Iron carbide 302, 529
Iron smelting 298303
direct reduction 306, 5289
Isobutane 496, 517
Isoflurane, see Fluranes
Isotopes 1368
JESSI semiconductor R&D programme
531
Joule Thompson effect 73, 75
Kinetic theory 1, 22, 27, 28
Kraft pulping 441
Krypton 483, 488, 490, 492
LAir Liquide 44, 245, 386
Langmuir isotherm 1023, 386
Laser 4912
Laser cutting 3323
Laser welding 333
Lavoisier, Antoine 2, 163
LCD glass coating 47980
LD, see Steelmaking, LindDonawitz
Le Chatelier 270
Leaks in pipework 2234
acid gases 279
helium leak detector 223
ultrasonic leak detector 224
valve stem seals 224
Leapfrog economies 59
LeBlanc process 4, 295
LEDs, semiconductor gases for 422
Legionella pneumophilia 367
Leidenfrost phenomenon 266, 445
non-stick cryogenic freezing 3589
LEL, see Lower explosive limit
Life cycle analysis 2934
Limelight 4
Linde AG 45
Linde,Carl von 4, 89
Line-pack gas storage 236
Liquefiers, cryogenic 22
Liquid Air Corp. 45
INDEX 565
Liquid crystal displays 47980
Liquid helium 253, 46870
Liquid hydrogen
ortho and para forms 1289
production 1279
rocket propellant 4724
solid hydrogen 473
transport 253
vehicles 523
Liquid nitrogen
can rigidification 364
concrete cooling 4589
deflashing 458
differential embrittlement 462
environmental cabinets 450
fog 466
food freezing 3528
ground-freezing 456
milling and grinding 357, 4613
pipe-freezing 4546
plastic processing cooling 4646
shrink fitting 4568
solvent recovery 45961
theatrical effects 466
tyre shredding 462
Liquid oxygen
explosive mixture with hydrocarbon 84,
93
rocket propellant 4724
Lower explosive limit 266
Lubricants 20911
Lung function mixtures 50910
Lyophilisation, see Freeze-drying
Magic gas for nuclear particle detectors
517
Magnetic alloys 429
Magnetic bearings 756
Magnetron manufacture 4279
Malaysia Oxygen Company 45
Manifolded cylinder pack 247
MAPP, see Methyl acetylene; Propylene
Margarine 368
Mass spectrometer 17681
atmospheric pressure ionisation 17980
Faraday cup 176
magnetic sector 177
membrane concentration MS 1901
multiplier detector 176
multipoint MS 190
problems 1789
quadrupole mass filter 1778
time-of-flight 1801
Materials testing
Charpy V notch 197
oxygen compatibility 209
Matheson Corp. 254
Matrix organisation of gas companies 46
Mature industrial economies 59
Maximum levels of toxic/asphyxiant gases
in air 264
MBU, see Mechanical break-up nozzle
MCP, see Manifolded cylinder pack
Mean free path 24
Mean time between failures 2856
Mechanical break-up nozzle, 497
Melter-gasifier 3023
Membrane electrolysis 1267
Membrane separation 62, 11117
asymmetric membranes 115
fibre membranes 11516
future developments 5213
PRISM process 115
selective membranes 11315
Mercaptan 124, 543, 546
Merchant gases 43
Mercury
adsorber 126, 297, 370
reduction gas analyser 186
Mercury lamp 490
Messer Griesheim GmbH 44, 50, 318
Metal casting 31011
Metal halide lamp 490
Metal powder production 3389
Metalceramic seals 429
Metallorganic vapour phase epitaxy, 4202
Methane 3940, 1245, 494, 533
hydrates 124, 525
Methanol 529
Methyl acetylene 123, 320
Milling, cryogenic 4613
Millipore Corp. 389
Minienvironments for semiconductor
processing 530
Mixing in gas cylinders 224, 1456
Mixtures
filling 145
homogeneity 146,
see also Diffusion
interpolation of properties 10
production 1447
safety 1445
MMST R&D group 531
Moisture
in gas cylinders 146
measurement 1835, 390
Molecular flow 23
Molecular sieve 1001
see also Pressure-swing adsorption
Moncrieff theory of smell and chemical
formula 545
Monopoly in gas production 47
Monsanto Corp. 115, 392
Montreal Convention 412
MOVPE, see metallorganic vapour phase
epitaxy
MRI, see Nuclear magnetic resonance
MS, see Mass spectrometer
MTBF, see Meant time between
failures
NDIR, see Gas analysis, non-dispersive IR
Neon
lamp 490
solid neon cryogenic coolant 471
Nimbus satellites 2889
Nippon Sanso Corp. 44, 254
Nitric oxide 278, 477, 504
Nitrogen 37
beer 364
cryogenic nitrogen generator 889
firefighting 4745
foam inerting 347
nitrogen tunnels in wafer fabs 52931
oil wells 3412, 526
PSA process 107
tyres 4756
see also Liquid nitrogen
Nitrogen dioxide 278, 477
Nitrogen generator 88, 94
Nitrogen trifluoride 409
Nitrous oxide
accidents 278
anaesthetic gas 5046
flames 320
mixture with oxygen 506
production 1334
propellant in food packaging 499
NMR, see Nuclear magnetic resonance
Non-stick cryogenic freezing 3589
Northey pump 71, 259
NO
x
, see Nitric oxide; Nitrogen dioxide
see also Environmental issues
NPSH 232
Nuclear magnetic resonance 46870
Nuclear particle detectors 51617
Nuclear particle physics 517
NUCOR 302, 529
Nusselt number 18
OOP, see Ozone depletion potential
Oil wells, injection of gas 3412, 524
On-site generation of gas 2545
see also PSA
Optical fibres 4225
Orbital welding 194, 327
Organometallics 4201
Orifice plate 13
Osaka Titanium Corp. 392
Osprey Metals Ltd 32930
Oxidants 274, 278
grease, oils 275
Oxy/acetylene cutting 317
Oxygen 37
accidents 2748
argon free 127, 487
breathing 5024, 51214
Brin process 4
ceramic firing 4835
chemical generation 118
compatibility testing 209
cryogenic oxygen generator 889
flame enhancement 30810, 4767
glass making 4767
maximum velocity in pipe 275
oxy-coal burner 310
paper-making 442
PSA process 108
purity in laser cutting 333
smelting 299303, 308
thermic lancing 322
under-flame lancing 30910, 4767
wastewater treatment 4325
see also Liquid oxygen
Oxygen candles, see Sodium peroxide
Oxygen concentrators (medical use) 109,
503
Oxygen index 2712
Oxygen masks for aviation 514
Oxygen/nitrogen ratio in economies 5960
Ozone 1301, 409, 431, 436, 438, 4401, 442
generators 1302
sterilisation 361
Ozone depletion potential 28990
CFCs/CFC replacements 290, 41112
Paper-making 4412
Particulates
contamination in gases 3746
measurement 407
plasma reactors 4067
Stakes formula 14
Passivation 390
Penning mixture 489, 490
Pentaborane 266
Perfluorocarbon 412
Perfluoropropane 510
Permeability, gases through polymers
11314
PET polymer pressure containers 237, 364
PETROX process for partial oxidation 343
PFC, see Perfluorocarbon
pH control with carbon dioxide 4389
Phase diagram 323
Phase separator 450
Phosphine 254, 3678, 397
Photochemical ozone creation potential
291
PID controller 220
Pilkington Company 479
Pilot-operated solenoid valve 220
Pipe-freezing 4546
Pipelines, long distance 252, 4712
Pipes
black iron pipe 192
copper pipe 192
INDEX 567
flexible bellows 193
joining pipes 1946
polyethylene 193
polymer hoses 1934
stainless steel 192
see also Pressure safety testing
Pipework design 202
Planck spectrum 321
Plasma coating 329
Plasma cutting 3256
Plasma enhanced CVD 409, 533
Plasma etching 4018
anisotropic 402, 4045
ashing of photoresist 4089
end-point detection 408
particulates 4067
problems 4058
selectivity 401, 404
theory 405
Plasma torch 3301
Plastic processing, cooling of 4647
PLC 220
Plethysmography 240, 509
Pneumatic controls 2201
Pneumatic lift of liquids 34851
Pocket calculator 58
POCP, see Photochemical ozone creation
potential
Poiseuilles equation 14
Poissons equation 19
Polymer gas containers 237
Prandtl number 18
PRAXAIR Corp. 44, 100, 107, 110
Pressure transfer of liquids 3489, 362
Pressure measurement 15961
Pressure parametric pumping 101
Pressure raising circuit 231
Pressure regulators, see Regulators
Pressure safety testing 197201
acoustic emission 2001
cracks 201
dye penetrant 198
hydraulic testing 8, 1978
magnetic 198
pneumatic testing 8
resistive 1989
ultrasonic echo 199200
X-ray/ -ray 199
Pressure vessels 1967
Pressure-swing adsorption 98
adsorber bed dynamics 1034
future development 5201
hydrogen 107
mass transfer zone 103
molecular sieves 1001
nitrogen 107
oil well nitrogen injection 3412, 526
oxygen 108
practical considerations 11011
pressure equalisation 106
single bed 110
small plant 1089, 503
wastewater treatment 434
Priestley, Joseph 2, 363
Printed circuit boards 427
PRISM membrane process 115
Product exchange 52
Production of gases 43
Propane 1245, 496
Proportional nuclear particle detector
51617
Propylene 1234, 320, 551
PSA, see Pressure-swing adsorption
PTFE 208, 209, 2712, 280, 359, 387
hose 193
oxygen index 271
Pulsation filter 203
Pumps, cryogenic 2313
gear pump 71
NPSH 232
see also compressors; pressure transfer
of liquids; Vacuum pumps
Purging 2789, 2812, 41314
Purifiers 13744
adsorption 13840
catalytic 1401
end-point detection 390
getter 1402
palladium diffusion 142
overheating 143
ultraclean 38990
Purity of gases
gaseous contamination 3767
paniculate contamination 3746
research 515
semiconductor production 3737
Pykrete 443
Pyrophoric 339
Quartz film 482
Quartz microbalance 1845, 390
Quartzhalogen lamp 4889
Radiation detectors, see nuclear particle
detectors
Radioactive isotopes 1367, 156, 487
Radon 263
Ramsay, Sir William 3
RanqueHilsch vortex tube, see Vortex
tube
Raoults law 36
Reactive ion etching 403
Recuperative burner 310, 476
Reformer gas 125
Refrigerants, see Chlorofluorocarbons;
Hydrochlorofluorocarbons
Regenerative burner 310, 476
Regulators 21213
Reversing heat exchanger 84
Reynolds number 16
RHE, see Reversing heat exchanger
RIE, see Reactive ion etching
Rocket propellants 4723
Roll-over phenomenon, 2345
Rootes blower 71, 259
RTP, see Semiconductors, rapid thermal
processing
Safety
burning leaks 273
containment of gases 265
cryogenics 2656
double pipe 26388
equipment design 27984
working in flammable atmospheres 272
see also Accidents; Intrinsically safe
electrics
Sand moulds for casting
binding 31011
SASOL process 529
Scaling of gas plants 536
future changes in scaling 520
Scarfing 319
Scheele 2, 129
SCR, see Selective catalytic reduction
Scrubbers
dry 2967
heated 2967
plasma 4123
wet 2956
see also Electrostatic precipitation;
Filters
SDS, see Gas cylinders, adsorber filled
Seals 678, 2069
active seals 678
compressors 678
elastomers 2078, 210
labyrinth 67
O-rings 2067
swelling in solvents 2078
thread sealant 209
Searls equation 26
Selective catalytic reduction 4778
Semiconductor industry 3713
Semiconductors 392416, 4202
annealing 413
crystal growth 392
CVD 3946
doping 3967
epitaxial layers 3934
hybrid circuits 41819
IC assembly 41618, 41920
oxide growth 3979
plasma processing 40111
point-of-use chemical generation 41516
rapid thermal processing 399400
sputtering 4001
wafer cleaning processes 41011, 41516
Shin-Etsu Corp. 392
Shrink fitting 4568
Siemens AG 392
Sieve trays 91
Silane 1356, 3945
Silica gel 99
Silicon 3924
Silicon carbide 534
Silicon strain gauge 161
Silicon tetrafluoride 409
Sintering 3336
alumina 334
ferrites 3345
see also High T
c
superconductors
Smell of gases 191, 5436
Smelting, use of oxygen 299303, 308
SMIF 374, 531
Smokestack economies 59
Soda-lime adsorber 508, 509, 512
Sodium lamp 490
Sodium peroxide 118
Soldering 4257
Solubility of gases in liquids 347
Solvay cell 126
Solvent recovery 45961
SORBEX process 101
Southampton University 443, 456
Spacecraft atmospheres 51314
Sparging 362, 440
Special gases 41
product exchange 52
production 1347
Specific impulse 4723
Speed of sound 549
Spray can 578, 496500
alternative can systems 498500
environmental impact 294, 498
gas storage 2367
Polygas can 499
Sputtering 4001, 4802
Stabilisation of drinking water 4367
Stable isotopes
extraction 137
Stainless steel
production 3067
Steelmaking 3037
basic oxygen processes 3045
future 5289
LindDonawitz steel-making 3045
mini-steelworks 3056
Stefan-Boltzmann radiation law 228, 3201
Stellite 322
Stenchants 124
Stirling cycle refrigerator 4524, 471
Stokes formula 14
Structured packing 92
Stythe 263
Sublimation 32
INDEX 569
Suction head, see NPSH
Sulphur dioxide 441
Sulphur hexafluoride 411, 476, 483
circuit breakers 4857
eye surgery 510
sublimation 32
Superconductors 46870, 517
Supercooling 444
Supercritical fluid 324
extraction 34, 227, 370, 371, 5278
Superheating 444
Superinsulation 228, 229
Sutherland formula 28
Swagelok 1956, 224
Synthetic petrol 529
Take-or-pay 52
TCD, see Gas analysis, katharometer
Tea-tree oil 367
Technology Foresight 518
Temperature measurement 1612
Temperature swing adsorption 84
TEOS, see Tetraethylorthosilicate
Tetraethylorthosilicate 395
Tetrahydrofuran 122
Tetronics Ltd 330
Texas Instruments Corp. 531
Thermal conductivity
gases 267
insulation materials 229
Thermal conductivity detector, see Gas
analysis, katharometer
Thermic lancing 322
Thiols, see Mercaptans
TI Chesterfield Ltd 48
TIG, see Arc welding, tungsten inert gas
TOP (MS), see Mass spectrometer,
time-of-flight
Tohoku University 391, 530
Tonnage gases 49
Transport of gases 2523
cylinders on vehicles 284
Triethyl aluminium 266
Trimethyl aluminium 422
Triple point 32
TSA, see Temperature swing adsorption
TSMC Taiwan Corp. 530
Tungsten hexafluoride 395
Tungsten working 334
Turbulent flow 1519
in liquid cryogens 15, 17, 449
in pipework 18
Tuyeres 299
Two-phase flow 449
UEL, see Upper explosive limit
Ultraclean gas technology 37792
continuous flow 3912
cryogenic distillation 3789
cylinders 391
electropolishing 380
filters 3878
gas control panels 3825
inert atmosphere for SMIF box 374
MCG fittings 382
oxygen passivation 3901
passivation 3801
pipework 37985
point-of-reaction purity 385
prepurifiers 378
purifiers 38990
simulation of contaminant migration
3867
valves 382, 3889
VCR fittings 3812
Ultrapure gases for research 51417
Ultraviolet coatings 5002
Union Carbide Corp. 44
Upper explosive limit 266
Uranium hexafluoride 112
Uranium separation 11213, 117
US Steel Corp 302, 529
UV coatings, see Ultraviolet coatings
Vacuum insulation 226
flask 27
pipework 44950
superinsulation 228, 229
Vacuum pumps 25662
cryopumps 4512
diffusion 2578
ejector 257
gas ballast 2612
Northey 25960, 412
operational pressure ranges 262
Rootes 259
rotary vane 25860
scroll 259
turbomolecular 259, 261
Vacuum swing adsorption 101, 108
see also Pressure swing adsorption
Vacuum systems 236
Valves 2036
ball 204
butterfly 205
check/relief 204, 2056, 230, 2834
fluidic diode 204
opening characteristic 2056
poppet 204
problems 203
ultraclean 3889
Van der Waals equation 9, 74
Vaporisers
anaesthetic 5078
cryogenic 233
Vapour pressures 2832
VCR 224, 382
Venturi 13, 257
Vertech process 438
VG Instruments Company 115
Viscosity 28, 29
VOCs/volatile organic compounds 2901,
461, 500
Voltaix Corp 421
Vortex tube 76
VSA, see Vacuum swing adsorption
Wacker Corp 392
Wastewater treatment 42935
industrial wastewater 438
oxygen injection 4325
Water, drinking 4367
Water hammer 202
Welding, see Flame welding;
Arc welding
Wind tunnels, transonic 5267
Wobbe index 317
Wormald International Corp. 474
Xenon flashlamp 490
Zener barrier 272
Zeolites 100, 108, 254
see also PSA

Industrial Gases

Hochgeladen von

Dokumentinformationen

Originalbeschreibung:

Copyright

Verfügbare Formate

Dieses Dokument teilen

Dokument teilen oder einbetten

Freigabeoptionen

Stufen Sie dieses Dokument als nützlich ein?

Sind diese Inhalte unangemessen?

Copyright:

Verfügbare Formate

Industrial Gases

Hochgeladen von

Copyright:

Verfügbare Formate

Industrial Gases

!"#$ &'() #*+)*+#,*'--. -)/+ 0-'*1

the transition of flow from fluid to molecular-type flow in small orifices

the rate of interdiffusion of gas molecules;

Nusselt number, hD/k: heat input/heat lost from conduction, relating to

Prandtl number, Cu/k: kinematic viscosity/thermal diffusivity, i.e. the

Grashof number, buoyancy force/viscous force, relating to

Whether convective heat transfer can take place depends on whether

Conductive heat transfer is the most readily calculated of the three

has a significant non-gas operation (see above);

is not R&D (research and development) intensive, with 2% or less of

has a significant distribution business: except for pipeline business with

has an average growth rate: although industrial gases is a mature industry

in purchasing, bulk buying reduces costs;

there is always a tendency to gold-plate equipment which normally has a

market distortions often affect the very largest pieces of equipment of a

reliability and back-up considerations;

when plant exceeds handling equipment maxima: if simple, inexpensive

axial compressors cannot achieve high enough pressures except in the

the accumulation of uncondensible light gases, neon and helium, which

complexities associated with the most efficient use of heat exchangers;

the need to avoid the accumulation of small quantities of hydrocarbons

Weeping: liquid streams from a plate to that below effectively shorting

controllable microporosity so homogeneous beds of known capacity can

small macropore/interpellet void volume: materials with a large volume

flow into the micropores of the absorber is not instantaneous;

the bed will tend to heat up as adsorption takes place;

zeolites have two levels of microporosity;

pellets will not in general be identical spheres (the usual assumption in

the bellows used are subject to damage with pressure differentials;

buoyancy gauges in small tanks with top access;

differential pressure (DP) gauge on tanks (this is not very accurate on

calibrated venturi meters;

load cells indicating tank weight.

A hot wire which, when it is exposed to carbon-containing gas, changes its

high speed: to achieve a reasonable analysis cycle time on mixtures of

drift free: slowly eluting components, with peak widths of up to several

Some gases have virtually identical mass/charge ratio molecular ions. CO

Oxygen reacts with residual carbon in tungsten filaments, producing CO

materials: a limited number of metals, mostly grades of steel, are accept-

structural design, calculation methods and record-keeping;

design details, such as welds and fittings;

inspection and testing;

corrosion and fatigue life;

conditions of use (pressure cycling, temperature, pressure, intended

contents and trace contaminants);

ancillary equipment, such as pressure relief valves.

longitudinal joints in cylinders are under higher stress than circumferen-

vessels must show a safety factor of 4 versus maximum design pressure

where connections to pipework, manholes or other devices pierce the

impact testing such as Charpy V notch at a low temperature may be

the need to clear personnel away from a radiography site is awkward in

off-line processing of film is time-consuming;

image quality is often limited by fogging (due to scatter of X-rays), grain

thicker samples need higher voltage X-ray tubes or cobalt-60.

on-line (pressurised) connection/disconnection facilities, as provided by

convenience in use: pipework should ideally be easily accessible for

regulations on pipe locations (fuel gases should not be run near to

ease of fabrication: anyone who has tried to weld or solder a pipe in an

pressure losses: an excessive number of sharp elbow bends, for example,

in liquefied gas installations, in addition perhaps to cryogenic problems,

leakage across seat

high actuation force

!"#$ &'() #+)+#,'--. -)/+ 0-'1